Multistage histories of incremental accumulation, fractionation, and solidification during construction of large subvolcanic magma bodies that remained sufficiently liquid to erupt are recorded by Tertiary ignimbrites, source calderas, and granitoid intrusions associated with large gravity lows at the Southern Rocky Mountain volcanic field (SRMVF). Geophysical data combined with geological constraints and comparisons with tilted plutons and magmatic-arc sections elsewhere are consistent with the presence of vertically extensive (>20 km) intermediate to silicic batholiths (with intrusive:extrusive ratios of 10:1 or greater) beneath the major SRMVF volcanic loci (Sawatch, San Juan, Questa-Latir). Isotopic data require involvement of voluminous mantle-derived mafic magmas on a scale equal to or greater than that of the intermediate to silicic volcanic and plutonic rocks. Early waxing-stage intrusions (35–30 Ma) that fed intermediate-composition central volcanoes of the San Juan locus are more widespread than the geophysically defined batholith; these likely heated and processed the crust, preparatory for ignimbrite volcanism (32–27 Ma) and large-scale upper-crustal batholith growth. Age and compositional similarities indicate that SRMVF ignimbrites and granitic intrusions are closely related, but the extent to which the plutons record remnants of former magma reservoirs that lost melt to volcanic eruptions has been controversial. Published Ar/Ar-feldspar and U-Pb-zircon ages for plutons spatially associated with ignimbrite calderas document final crystallization of granitoid intrusions at times indistinguishable from the tuff to ages several million years younger. These ages also show that SRMVF caldera-related intrusions cooled and solidified soon after zircon crystallization, as magma supply waned. Some researchers interpret these results as recording pluton assembly in small increments that crystallized rapidly, leading to temporal disconnects between ignimbrite eruption and intrusion growth. Alternatively, crystallization ages of the granitic rocks are here inferred to record late solidification, after protracted open-system evolution involving voluminous mantle input, lengthy residence (105–106 yr) as near-solidus crystal mush, and intermittent separation of liquid to supply volcanic eruptions. The compositions of the least-evolved ignimbrite magmas tend to merge with those of caldera-related plutons, suggesting that the plutons record nonerupted parts of long-lived cogenetic magmatic systems, variably modified prior to final solidification. Precambrian-source zircons are scarce in caldera plutons, in contrast to their abundance in some peripheral waning-stage intrusions of the SRMVF, implying dissolution of inherited crustal zircon during lengthy magma assembly for the ignimbrite eruptions and construction of a subvolcanic batholith. Broad age spans of zircons (to several million years) from individual samples of some ignimbrites and intrusions, commonly averaged and interpreted as “intrusion-emplacement age,” alternatively provide an incomplete record of intermittent crystallization during protracted incremental magma-body assembly, with final solidification only when the system began to wane. Analyses of whole zircons cannot resolve late stages of crystal growth, and early growth in a long-lived magmatic system may be poorly recorded due to periods of zircon dissolution. Overall, construction of a batholith can take longer than recorded by zircon-crystallization ages, while the time interval for separation and shallow assembly of eruptible magma may be much shorter. Magma-supply estimates (from ages and volcano-plutonic volumes) yield focused intrusion-assembly rates sufficient to generate ignimbrite-scale volumes of eruptible magma, based on published thermal models. Mid-Tertiary processes of batholith assembly associated with the SRMVF caused drastic chemical and physical reconstruction of the entire lithosphere, probably accompanied by asthenospheric input.

Recent geochronologic and petrologic studies have convincingly demonstrated that large magma bodies, which form granitoid crustal plutons in Cordilleran-arc settings, were assembled incrementally and crystallize over 105–106 yr intervals (Coleman et al., 2004; Matzel et al., 2006a; Memeti et al., 2010; Frazer et al., 2014). Building on twentieth-century discussions (such as Daly, 1914; Kennedy and Anderson, 1938; Buddington, 1959; Smith, 1979; Lipman, 1984; Macdonald and Smith, 1988), these results have promoted renewed controversy concerning connections between volcanic and intrusive processes, especially how magma reservoirs for large ignimbrite eruptions are related to the emplacement of granitic plutons and the crustal depths at which silicic compositions are generated (Glazner et al., 2004; Metcalf, 2005; Bachmann et al., 2007b; Lipman, 2007; Miller et al., 2011; Davis et al., 2012; de Silva and Gregg, 2014; Frazer et al., 2014; among others). For example, geochronologic data and chemical patterns of volcanic and shallow plutonic rocks have been interpreted by some as indicating melt generation in the deep crust with minimal differentiation at shallow crustal levels, leading to pluton assembly in small increments that crystallized rapidly, with temporal and geometric disconnects between ignimbrite eruption and intrusion growth (Glazner et al., 2004; Bartley et al., 2005; Annen, 2009; Coleman et al., 2012; Zimmerer and McIntosh, 2012a; Mills and Coleman, 2013). Alternatively, others have concluded that rhyolitic magmas are commonly generated at low pressure (e.g., Tuttle and Bowen, 1958; Lipman, 1966; Fowler and Spera, 2010; Gualda and Ghiorso, 2013) by crystal-liquid separation in the upper crust, complemented by voluminous underlying cumulates (e.g., Hildreth, 1981, 2004; Bacon and Druitt, 1988; Vazquez and Reid, 2002; Bachmann and Bergantz, 2004; Deering et al., 2011). These and other recent studies (e.g., Wilson and Charlier, 2009; Allan et al., 2013; Pamukcu et al., 2013; Wotzlaw et al., 2014; Cashman and Giordano, 2014) have inferred broadly lenticular shapes, rapid assembly rates, and brief life spans for the shallow magma bodies that erupt as large ignimbrites. From complementary perspectives, this paper evaluates the vertical extent and temporal construction of the overall crust-mantle magmatic system inferred to have developed concurrently with ignimbrite eruptions.

Although available analytical methods and resulting data have thus far been only partly successful in addressing such issues, recent volcanologic, geophysical, petrologic, geochronologic, and modeling results for the Middle Tertiary Southern Rocky Mountain volcanic field (SRMVF) are consistent with prior proposals that large long-lived silicic volcanic fields are surface expressions of composite upper-crustal magma bodies comparable to the Sierra Nevada or Boulder batholiths (Smith 1960; Hamilton and Myers, 1967; Lipman, 1984, 2007; Bachmann et al., 2007b). Such volcanism offers sequential snapshots of processes during early magmatic evolution in continental arcs, while upper-crustal plutons provide a composite record of lengthy assembly and later crystallization.

Building on a prior review of incremental assembly and prolonged consolidation in Cordilleran magma chambers (Lipman, 2007), new data for the SRMVF, summarized here within a framework of global perspectives on continental-margin arc magmatism, permit more quantitative assessment of alternative geometric and genetic models for large silicic magma bodies, relations of mineral-crystallization ages to pluton-assembly processes, and resulting implications for magma-supply rates during evolution of the SRMVF and comparable Cordilleran-arc systems. Geological and geophysical data are used to estimate vertically extensive geometry and volumes of magmatic systems at the SRMVF, where mantle basalt recurrently mixed with lower-crustal melts to erupt thick sections of lavas and ignimbrites, underlain by largely coeval upper-crustal granitic batholiths with deep roots of more mafic residua. Petrologic features and crystallization ages of spatially associated SRMVF volcanic and granitoid rocks are interpreted to record late solidification after prolonged histories of open-system evolution in the shallow crust, involving recurrent magmatic recharge, lengthy (105–106 yr) residence as near-solidus crystal mush, and intermittent upper-crustal separation of liquid to supply eruptions. Ignimbrite eruptions in the SRMVF are inferred to record relatively brief episodes of increased mantle-magma recharge and concurrent upper-crustal pluton construction at focused sites within the broader areal extent of the volcanic field. The eruptible magmas that sourced SRMVF ignimbrites are considered to be short-lived and volumetrically minor, shallow portions of much longer-lived and vertically extensive underlying plutonic systems dominated by near-solidus mushy magma. Such temporal and volumetric features of volcanic-plutonic evolution at the SRMVF are suggested to be representative of continental-arc magmatism worldwide.

The mid-Tertiary SRMVF, site of 25 large ignimbrites (mainly 37–27 Ma) with related calderas and subvolcanic intrusions (Fig. 1; Table 1), provides an exceptional laboratory for processes of Cordilleran magmatism (Steven, 1975; McIntosh and Chapin, 2004; Lipman, 2007). In places, virtually pristine volcanic morphology has been exhumed by recent erosion; elsewhere, rugged topography and structural tilting expose multikilometer volcanic sections, down into upper levels of subvolcanic intrusions. Small granitoid plutons, many spatially and temporally associated with ignimbrite calderas, are exposed at near-roof level (Table 2), while geometry and composition of a vast composite batholith that is vertically extensive beneath the volcanic locus are constrained by geophysical and geochemical modeling.

As summarized more fully previously (Lipman, 2007, and references), dominantly intermediate-composition lavas and associated breccias (andesite, dacite) were voluminous precursors to most ignimbrite eruptions, and eruption of similar lavas continued concurrently with the major ignimbrites, commonly filling caldera depressions (Steven and Lipman, 1976). At the San Juan locus, the central volcanoes and their clastic aprons (∼25,000 km3) constitute almost two thirds of total volcanic volume (Lipman et al., 1970, 1978). Basalt is nearly absent, despite repeated searches for primitive compositions. Mafic alkalic dikes (lamprophyres) with mantle isotopic signatures locally intrude pre-Tertiary rocks in the northern San Juan Basin (Gonzales et al., 2010; Lake and Farmer, 2015; Gonzales and Pecha, 2015), but such compositions have not been identified centrally within loci of large-volume San Juan volcanism. Volcanic centers tended to migrate from north to south in the SRMVF, both intermediate-composition lava eruptions and ignimbrites (Fig. 1B, Table 1; Lipman, 2007); the general southward migration is parallel to that long documented for eruptive centers in the Basin-Range region, probably related to disruption of the subducted Farallon plate (Stewart et al., 1977; Lipman, 1980; Henry and John, 2013).

Structural unroofing, associated with later extension along the Rio Grande rift and deep erosion of this high-standing region, has exposed broadly synvolcanic batholithic intrusions associated with the ignimbrite centers. The earliest well-documented regional ignimbrite, erupted from a caldera source in the SRMVF, was the far-traveled Wall Mountain Tuff at 37 Ma (Chapin and Lowell, 1979; Zimmerer and McIntosh, 2012a), erupted from the Princeton batholith area in the Sawatch Range (Fig. 1); the southernmost, and among the youngest ignimbrites, was eruption of the Amalia Tuff from the Questa caldera at 25 Ma in northern New Mexico (Lipman, 1988; Tappa et al., 2011; Zimmerer and McIntosh, 2012b). Farther north in Colorado, the mid-Tertiary magmatism is marked by scattered shallow intrusions and sparse small erosional remnants of lava and tuff. A thick section of welded tuff associated with an intrusive complex in the Never Summer Mountains may record remnants of a small isolated caldera system active at 28–29 Ma (O’Neill, 1981; Jacob et al., 2011, 2015), concurrent with peak activity in the San Juan Mountains to the south.

Geographically and temporally between the early and late centers, the San Juan region contains the largest preserved erosional remnant of the composite Oligocene volcanic field (Larsen and Cross, 1956; Steven et al., 1974). The San Juan locus is notable for the large number of high-volume, compositionally diverse ignimbrites (cumulatively, ∼15,000 km3) and associated caldera collapses, at least 18 in the 3 m.y. interval 30.1–26.9 Ma (Table 1). Unzoned uniform crystal-poor rhyolite, crystal-rich dacite (“monotonous intermediates”), and ignimbrites that grade from initially erupted rhyolite upward into dacite are present in subequal numbers. Sizable precursor Plinian-fall deposits have not been recognized beneath any of these ignimbrite types, contrary to some recent inferences (e.g., Gregg et al., 2012; Cashman and Giordano, 2014).

The composite SRMVF, now widely erosionally dissected, was one site of discontinuous Middle Tertiary Cordilleran magmatism, continuing southward through the Mogollon-Datil region in New Mexico (Elston, 1984; Ratté et al., 1984; McIntosh et al., 1992), into Trans-Pecos, Texas (Henry and Price, 1984), and the vast Sierra Madre Occidental of northern Mexico (McDowell and Clabaugh, 1979; Ferrari et al., 2007; McDowell and McIntosh, 2012). The SRMVF was originally comparable in size, composition, and magmatic duration to large young ignimbrite terranes such as the well-documented Altiplano-Puna volcanic complex (APVC) of the central Andes (de Silva, 1989; Lindsay et al., 2001; Schmitt et al., 2002; de Silva and Gosnold, 2007; Salisbury et al., 2011; del Potro et al., 2013).

Within the SRMVF, the San Juan magmatic locus provides an exceptional natural laboratory for evaluating the geometry, composition, and emplacement history of intrusive bodies in relation to broadly associated surface volcanism: (1) Volcanic rocks are widely preserved but eroded to depths that expose shallow parts of contemporaneous granitoid intrusions. (2) Detailed regional volcanic stratigraphy and abundant petrologic, geochemical, and geochronologic data provide a comprehensive record of eruptive history. (3) Rocks of the San Juan locus were emplaced mainly onto the structurally simple Colorado Plateau block that has been broadly stable since craton formation, thereby permitting well-constrained gravity and seismic modeling of subsurface intrusion geometry. (4) Isotopic contrasts between Precambrian crust and underlying mantle provide robust geochemical tracers for evaluating magma-generation processes.

The vertical extents and volumes of large subvolcanic batholiths and their component plutons, even if only broadly determined, provide critical parameters for interpreting geochronologic and petrologic data on magma-crystallization history, duration of magma-body assembly, and magma-supply rates during volcanism. Upper-crustal exposures document that many Cordilleran plutons in the western United States reached shallow levels, some intruding broadly cogenetic volcanic deposits (Buddington, 1959; Smith, 1960; Hamilton and Myers, 1967; Lipman, 1984). Such granitic bodies have been increasingly recognized as emplaced incrementally during protracted intervals (e.g., Pitcher, 1979; Wiebe and Collins, 1998; Coleman et al., 2004; Matzel et al., 2006a; Miller et al., 2011; Davis et al., 2012; Mills and Coleman, 2013). Many recent models infer assembly and solidification of silicic plutons as successions of stacked sill-like bodies with overall broadly lenticular shapes (Cruden, 1998; Petford et al., 2000; Glazner et al., 2004; Bartley et al., 2008; Annen, 2009). Estimates of pluton thickness have varied widely, in part because of limited topographic relief in even the most rugged terrains, and many large granitic plutons have been inferred to be 10 km or less thick (Hamilton and Myers, 1967, 1974; Cruden et al., 1999; McNulty et al., 2000; Annen, 2009). However, geologic-mapping and geobarometric data for individual plutons, deep-crustal exposures through tilted magmatic arcs, and geophysical data for crustal structure beneath large volcanic fields increasingly provide robust evidence for vertically extensive subvolcanic batholiths and their component plutons (Saleeby et al., 2003; Ducea et al., 20010; DeBari and Green, 2011; Jagoutz and Schmidt, 2012). Such bodies are here interpreted to occupy much of the crust and to be associated with large-scale compositional modification of the underlying lithospheric mantle.

San Juan Batholith: Geophysical Expression

Deep Bouguer gravity lows (to –340 mGal) in the Southern Rocky Mountains (Fig. 2A), which have long been interpreted as expression of upper-crustal Tertiary granitic intrusions, coincide spatially with major clusters of ignimbrite calderas in the San Juan region (Plouff and Pakiser, 1972; Drenth et al., 2012), with several older calderas and subvolcanic intrusions along the Sawatch Range (Fig. 2B; Isaacson and Smithson, 1976; Case and Sikora, 1984; McCoy et al., 2005), and with the Questa caldera and multiple associated granitic plutons exposed in the Latir Range (Cordell et al., 1985; Lipman, 1988).

The gravity anomaly associated with the San Juan volcanic locus has been documented and modeled in particular detail; it has steep marginal gradients and a subdued interior structure; most individual calderas lack expression, and the lowest gravity values are within a central area where calderas are absent. Such features led to the interpretation that the bulk of the anomaly is due to the presence of a large composite batholith: ∼10,000 km2 in area, with an average density contrast to country rock of 100 kg/m3, its top widely occurring at a few kilometers below present-day surface, and extending to depths greater than 19 km (Plouff and Pakiser, 1972). As modeled in elegant detail, combining more abundant gravity data with seismic-velocity profiles and adjusting for low densities of near-surface rocks, the San Juan batholith has more recently been inferred to underlie an area of ∼8200 km2 beneath central parts of the volcanic cover (Drenth et al., 2012). Using a preferred density contrast of 80 kg/m3 between assumed densities for the Tertiary batholith (2620 kg/m3) and upper-crustal Precambrian basement (2700 kg/m3), these authors obtained an average batholith thickness of 13 km and a volume of 82,000–130,000 km3 (Figs. 3A–3B).

Compositions and densities of exposed San Juan granitoid intrusions suggest, however, that the batholith volume and thickness may be even larger. By comparison with measured densities for granitic rocks elsewhere such as from the Sierra Nevada batholith (Oliver et al., 1993; Moore, 2000, his fig. 6.20), the relatively low assumed density (2620 kg/m3) modeled for the batholithic rocks in the San Juan region by Drenth et al. (2012) would imply an average composition (∼75% SiO2) of silicic granite (Fig. 3C). Few reliable density measurements exist for exposed Tertiary intrusions in the San Juan region, but the largest bodies and most common compositions are granodiorite (∼62%–68% SiO2). Intrusive rocks with >70% SiO2 are rare, relatively small, mainly peripheral to the geophysically defined batholith, and mostly emplaced late during growth of the SRMVF (fig. 8 inLipman, 2007; Gonzales and Pecha, 2015). Granodiorite to quartz monzonite average compositions and accompanying higher densities are also typical for exposed levels of Cordilleran plutons and batholiths elsewhere in the western United States: e.g., 2690 kg/m3, 68% SiO2 for the Sierra Nevada (Oliver et al., 1993), and 2660 kg/m3 for the Boulder batholith (Biehler and Bonini, 1969; Vejmelek and Smithson, 1995). The average upper crust is also granodioritic (Taylor and McClennan, 1981; Rudnick and Gao, 2003).

If the average composition of the San Juan batholith were 68% SiO2, with a density of 2660 kg/m3 (Fig. 3) and a lesser contrast of 60 kg/m3 with slightly denser country rocks, the average batholith thickness would be 20 km or more. The overall batholith could be even thicker if the average batholith composition were less silicic (as suggested by the exposed intrusions), or if the density contrast at roof level decreased to near zero as granitoids became more mafic at depth. Alternatively, the country rocks around the San Juan batholith could be more dense (perhaps 2750 kg/m3), while maintaining the inferred contrast of 80 kg/m3 (Fig. 3B). A dense upper crust may be inconsistent, however, with the regionally high elevations and near-constant Moho depths (Hansen et al., 2013) in the Southern Rocky Mountains beyond the batholith area. More complex geometric models that could fit the gravity data include a multilayered batholith, becoming more dense downward, as expected from petrology and exposed crustal sections (see following sections).

In either case, if the San Juan batholith is less silicic in average composition than implied by the density model of Drenth et al. (2012) and/or becomes more mafic at depth, the geophysically modeled thickness of granitic rocks beneath the San Juan volcanic locus would necessarily be greater. More mafic compositions have been inferred to occur commonly at deeper levels in subvolcanic magma bodies, based on eruptions that record compositional gradients (e.g., Hildreth, 1981; Bachmann and Bergantz, 2004). Consistent with such interpretations from the volcanic record, some Cordilleran granitic plutons expose mafic lower zones (Best, 1963; Coleman et al., 1995; Sisson et al., 1996; Wiebe et al., 2002; Miller et al., 2011; Putirka et al., 2014). Many discussions of magma geometry and intrusion depth note that geophysical anomalies provide few constraints on extent below depths where density and seismic-velocity contrasts diminish between intrusion and wall rocks.

Several early San Juan calderas, notably the Platoro complex (30.1–28.6 Ma) to the southeast and the Bonanza-Marshall cluster (34–33 Ma) to the northeast (Fig. 2A), lie beyond the batholith margins as modeled by Drenth et al. (2012), hinting at additional interpretive complexities. Bonanza, where resurgent intrusions range from mafic granodiorite to high-silica aplitic granite (Lipman et al., 2013), has modest gravity expression and was included as an outlier of the batholith as modeled by Plouff and Pakiser (1972). The Platoro complex is geophysically obscure despite having sourced five large-volume dacitic ignimbrites, each followed by eruptions of andesitic lavas, and associated with widely scattered intrusions mainly of mafic granodiorite (Lipman, 1975; Dungan et al., 1989; Lipman et al., 1996). The locations of these calderas, peripheral to the geophysically defined batholith, along with relatively mafic associated ignimbrites, caldera-filling lavas, and late granitic intrusions, suggest that any associated upper-crustal pluton lacks significant density contrast with host rocks. These relations further point to the probability that the overall San Juan batholith, including such mafic components, is likely to be substantially larger than that imaged and modeled from available gravity and other geophysical data.

Based on these interpretations of the San Juan geophysical data, comparisons with other Cordilleran plutons and with exposed crustal sections through volcanic arcs (reviewed in a following section), the granitoid San Juan batholith is here inferred to have its roof ∼2–5 km below the present-day surface, an average thickness of at least 20 km (more likely 25–30 km), and thereby occupying much of the Southern Rocky Mountains crust (thickness 40–45 km; Prodhel and Lipman, 1989; Hansen et al., 2013). The companion batholith beneath the Sawatch Range to the north (Isaacson and Smithson, 1976; Case and Sikora, 1984; McCoy et al., 2005) and smaller bodies associated with the Questa caldera in northern New Mexico (Cordell et al., 1985; Lipman, 1988) appear to have broadly similar density structure and vertical dimensions; these also likely extend to deep crustal levels. Because gravity and seismic data are mainly sensitive to crustal layering, geophysical models have limited capacity to distinguish aggregates of steeply dipping plutons as the building blocks of crustal batholiths (e.g., fig. 4 inSaleeby et al., 2003; fig. 3 inPaterson et al., 2011).

Sources and Age of the San Juan Geophysical Anomaly

The correlation between the gravity anomaly and caldera sites of large-scale silicic eruptions (Fig. 2A) suggests that the volumetric bulk of the San Juan batholith was assembled mainly during the ignimbrite eruptions and associated caldera formation: at 28.8–26.8 Ma (Table 1), or more broadly ∼34–27 Ma if the Marshall and Bonanza caldera cycles that lie just outside the intrusion modeled by Drenth et al. (2012) are included. Granitoid intrusions of varying size and composition are exposed within eroded calderas (Fig. 2A; Table 2), but the absence of geophysical expression for individual plutons or the overlying calderas indicates that granitic rocks are widespread beneath the clustered calderas and intervening areas.

No sizable contribution to the gravity anomaly seems likely from the granitic rocks of the Proterozoic basement; such rocks are distributed widely in southwestern Colorado (Tweto, 1979), far beyond margins of the San Juan gravity anomaly without obvious geophysical expression (Drenth et al., 2012, p. 674). Late Cretaceous to Paleocene intrusions (75–60 Ma; “Laramide”) associated with the northeast-trending Colorado mineral belt (Cunningham et al., 1994; Stein and Crock, 1990; Chapin, 2012) also appear to have been unimportant: The San Juan gravity anomaly is elongate east-west, at a high angle to the mineral belt trend (Fig. 2A), and the small intrusions of Laramide age southwest of the San Juan region (e.g., La Plata, Rico) lack large-scale geophysical expression.

Small stocks and laccoliths associated with the central volcanoes of dominantly intermediate compositions (ca. 34–30 Ma) that preceded the ignimbrite eruptions also seem unlikely to record major mid- to upper-crustal batholith growth. These volcanoes are scattered widely beyond the area of the main gravity anomaly (Steven et al., 1974; Tweto, 1979) but lack geophysical expression indicative of sizable silicic intrusions. Any larger intrusions at depth, associated with preserved central volcanoes, would have to be relatively mafic (diorite?) and/or relatively deep to minimize density contrast. The early magma fluxes would still have been high, because the composite volume of central volcanoes was large (greater than that of ignimbrites at the San Juan locus; Lipman et al., 1970). Additionally, intrusive fluxes may have been higher and intrusions larger at early central volcanoes within the area of the gravity anomaly (though exposed remnants of such proximal volcanoes are not obviously larger or more silicic than peripheral ones). Such waxing-stage intrusion and volcanism (Colucci et al., 1991) likely warmed the upper crust, providing needed thermal preparation for the ignimbrite eruptions and associated assembly of multiple upper-crustal plutons into a composite batholith (Lipman et al., 1978; Jellinek and DePaolo, 2003; de Silva and Gosnold, 2007; Gregg et al., 2012; Gelman et al., 2013).

Scattered granitic bodies were also intruded widely in the San Juan region during protracted waning magmatism (ca. 26–10 Ma) but again seem unlikely to have contributed significantly to batholith growth. Such intrusions are typically small, many as dikes, sills, and modestly larger laccoliths. Most are exposed peripherally to the geophysically defined batholith and associated calderas, with major concentrations west and south of erosionally preserved volcanic rocks (Gonzales and Pecha, 2015). Many of these are silicic representatives of the bimodal magmatic suite associated with diffuse regional extension and opening of the Rio Grande rift zone (Christiansen and Lipman, 1972; Chapin, 1979), in contrast to the dominantly intermediate-compositions and arc-type petrology of the Oligocene magmatism. The few caldera-related loci of waning magmatism are peripheral to or beyond the geophysically defined San Juan batholith. Small intrusions of silicic rhyolite intruded ring fractures of the 28.3 Ma Uncompahgre caldera at 19 Ma in the Lake City area; small 20 to 22 Ma rhyolites adjacent to the Platoro caldera complex are similarly millions of years younger and compositionally more evolved than caldera-forming events there.

SRMVF Magma Generation and Volume

Diverse petrologic and geochemical studies on processes of magma generation for the SRMVF and similar Cordilleran igneous provinces have led to broad consensus: The dominant andesitic to rhyolitic magmas were generated by rise of voluminous mantle-derived basalt that provided heat to assimilate variable amounts of lower crust, as the evolving magmas crystallized and fractionated (e.g., Lipman et al., 1978; DePaolo, 1981; DePaolo et al., 1992; Hildreth and Moorbath, 1988; Johnson, 1991; Riciputi et al., 1995; Farmer et al., 2008; Kay et al., 2010; Jacob et al., 2015). The much-discussed potential for mantle-generated mafic input to rejuvenate and prolong the life spans of upper-crustal magmatic systems (e.g., Smith, 1979; Hildreth, 1981; Mahood, 1990; Bachmann et al., 2007b; Cooper and Kent, 2014; de Silva and Gregg, 2014) has been explicitly proposed to explain complex petrography and mineral chemistry of large ignimbrites in the SRMVF and elsewhere (e.g., Lipman et al., 1997; Bachmann et al., 2002; Bachmann and Bergantz, 2004; de Silva and Gosnold, 2007; Huber et al., 2012).

The volumetrically dominant intermediate-composition lavas (andesite-dacite) of the SRMVF contain diverse crystal cargos indicative of crystallization at variable depths and volatile contents in the upper crust (clinopyroxene vs. amphibole assemblages; Colucci et al., 1991; Parat et al. 2005). Dacites are typically crystal rich; disequilibrium phenocryst textures (complex zoning and resorption) are common; and groundmass compositions are rhyolitic (Lipman et al., 1997; Bachmann et al., 2002). In contrast, most erupted rhyolites are crystal poor, and phenocrysts are in near-equilibrium with groundmass (absence of intricate zoning or resorption textures). Bulk chemistry, groundmass compositions, and phenocryst assemblages (sodic sanidine, absence of quartz) are indicative of low-pressure fractionation in the upper crust (Lipman et al., 1978; Bachmann and Bergantz, 2004; Huber et al., 2012; Gualda and Ghiorso, 2013), rather than being inherited from melting of lower-crustal sources as proposed by others (Annen et al., 2006; Tappa et al., 2011; Coleman et al., 2012). Some large crystal-poor ignimbrites are uniform rhyolite; others are zoned upward into more crystal-rich dacite (Table 1), suggesting transition to additional crystalline residua deeper in a vertically extensive source body. For large ignimbrites in the SRMVF and elsewhere, caldera areas and subsidence depths indicate that the erupted magmas were laterally extensive bodies only a few kilometers thick. Probably most available crystal-poor silicic magma was erupted, down to a “viscosity barrier” (Smith, 1979; Karlstrom et al., 2012), but late-erupted scoria of crystal-rich mafic dacite and andesite provide direct samples of mafic cumulates in compositionally zoned ignimbrite eruptions such as the Bonanza and Carpenter Ridge Tuffs (Lipman et al., 2013; Bachmann et al., 2014). Geochemical signatures of crystal accumulation complementary to rhyolite fractionation probably are obscured in many late-erupted ignimbrite dacites (and associated plutons) by high volumetric ratios of the source magma mush relative to erupted evolved liquid (Bachmann and Bergantz, 2008; Deering et al., 2011; Gelman et al., 2014).

Because the >60,000 km3 volume of eruptions in the SRMVF (Lipman, 2007) was accompanied by much greater (but less constrained) volumes of associated granitoid intrusions, perhaps on the order of 250,000–300,000 km3 (Table 3; Fig. 4), the volume of associated mantle-generated mafic magma becomes critical for models of crustal and lithospheric evolution. Several attempts to estimate proportions and total volumes of required mantle basalt, based on compositional and isotopic mass-balance calculations, suggest 1:1 to >2:1 ratios of mafic magma to the intermediate to silicic magmas that reached upper-crustal levels (Riciputi and Johnson, 1990; Perry et al., 1993; Farmer et al., 2008; Kay et al., 2010; Lake, 2013). For just the San Juan magmatic locus, such geochemical modeling suggests that 200,000–300,000 km3 of mantle basalt hybridized the preexisting crust to generate the Middle Tertiary volcanism and underlying batholith; total magma volume for the San Juan locus would be ∼400,000–500,000 km3 (Table 3; Fig. 4). Volcanic deposits and the batholith are estimated at about half of this volume, with mantle magma and crustal melt in subequal amounts, and the remaining half of total volume present as deeper residue (cumulates, restite) from generation of the crustal magmas. From such a perspective, the ignimbrite “supereruptions” are only a few percent of the total magmatic system.

Batholith Growth and Crustal Structure

The prodigious inputs of basalt involved in generating the SRMVF and associated batholith might suggest substantial thickening of the Rocky Mountain crust (Riciputi and Johnson, 1990), possibilities that can be evaluated by geophysical data for the present-day crustal structure. Somewhat surprisingly, the crustal thickness in the Southern Rocky Mountains (∼40–45 km) is no greater than that of the adjacent High Plains (Prodehl and Lipman, 1989). Elevation even tends to correlate inversely with Moho depth, without the presence of any deep crustal root to support the mountain topography (the “rootless Rockies”: Hansen et al., 2013). Rather, the high elevations are gravitationally supported by atypically low densities (low seismic velocities) within the middle to lower crust and in the upper mantle. As a result, if the inferred granitoid thickness (20–30 km) for the San Juan batholith and volume of associated mafic input associated with the SRMVF are valid, space in the lower crust is inadequate to accommodate the volume and thickness of the dense (high-velocity) mafic residue (cumulates, restite from partial melting) that would be complementary to generation of the San Juan and other batholith loci beneath then SRMVF. Because seismic velocities in the lower crust beneath the SRMVF are relatively slow, much of the expected mafic residue must lie beneath the seismic Moho and/or have delaminated (see later discussion).

Comparisons with Other Cordilleran Volcanic Regions

The inferred vertically extensive geometry of the composite SRMVF batholith is suggested to be a well-constrained representative example of volcano-plutonic connections. Analogous correlations between upper-crustal gravity lows and caldera-forming volcanism in other Cordilleran volcanic regions have been similarly interpreted as evidence for shallow batholithic intrusive complexes in diverse structural settings. Examples include the Latir and Mogollon-Datil fields farther southwest in New Mexico (Cordell et al., 1985; McIntosh et al., 1992; Schneider and Keller, 1994), the Marysvale volcanic field in Utah (Steven et al., 1984), Elkhorn Mountains volcanic field and underlying Boulder batholith in Montana (Robinson et al., 1968; Hamilton and Myers, 1974; Biehler and Bonini, 1969; Vejmelek and Smithson, 1995), and the Yellowstone caldera cluster (Eaton et al., 1975; Christiansen, 2001).

A particularly informative comparison with the San Juan region is the APVC of the central Andes (de Silva, 1989; de Silva and Gosnold, 2007; Kay et al., 2010). The APVC is a younger (1–10 Ma) but otherwise striking analog to the SRMVF, in terms of volcanic compositions, volumes, eruptive processes, and magmatic duration (table 4 inLipman and McIntosh, 2008). No granitic intrusions are exposed at the little-eroded APVC, but combined seismic, gravity, and deformation data (Chmielowski et al., 1999; Zandt et al., 2003; Fialko and Pearce, 2012; Ward et al., 2014) document a large subvolcanic batholith. Seismic attenuation in the middle to upper crust is inferred to define a zone of mushy magma at depths of 4–25 km below sea level that is coextensive with the volcanic field (Ward et al., 2014). An associated gravity anomaly (–50 mGal) has been modeled as evidence of an extensive batholith from near the surface to 30 km depth or more (del Potro et al., 2013). The mush zone has been interpreted to be as much as 200 km across with a volume of ∼300,000 km3, suggesting a ratio of batholith to erupted magma (∼15,000 km3) of 20:1–35:1 (Ward et al., 2014), i.e., notably higher than often-cited estimates of 10:1 or less (Crisp, 1984; White et al., 2006) or ratios calculated here for the SRMVF (Table 3).

Because the batholith geometry inferred beneath the SRMVF (mosaic of vertically extensive intrusions, collectively 20–30 km thick, likely becoming more mafic downward; residue of mafic cumulate gabbro and restite at depth) comes mainly from geophysical and petrologic data that are open to alternative interpretations, relations observable more directly from exposed tilted plutons and sections through deep parts of magmatic arcs elsewhere provide additional evidence for thick subvolcanic Cordilleran batholiths.

Thickness of Mid- to Upper-Crustal Plutons

Although relatively rarely exposed, partial sections through several tilted plutons and batholiths display original thicknesses of 10 km or more, providing direct evidence for the thick-batholith geometry and dominantly granodioritic compositions (becoming more mafic at depth) inferred from roof-zone intrusions in the SRMVF.

Along the Nevada-Arizona boundary (Colorado River extensional corridor), several tilted Miocene plutons (Spirit Mountain, Searchlight, Aztec Wash) provide partial sections of lithologically diverse granitic rocks, 5–10 km thick, including roof contacts (Miller and Miller, 2002; Walker et al., 2007; Miller et al., 2011). These plutons are compositionally diverse, most containing relatively homogeneous granite that overlies and merges laterally and with depth into less-evolved granodiorite; heterogeneous zones of alternating mafic and felsic sheets provide evidence for recurrent open-system mafic recharge.

At the Wooley Creek pluton in northern California, geologic and petrologic evidence indicates that erosion has exposed a sequence of cogenetic granitic rocks at least 12 km thick. No base is exposed for these rocks, which crystallized at 159–158 Ma and are zoned downward from granite to tonalite (Barnes et al., 1986; Coint et al., 2013).

In the Coastal batholith of the Northern Cascades and southern British Columbia, elongate (“tadpole”) plutons dominantly of granodiorite, separated by metamorphic septa, display textures indicative of depth zones from mesozonal to hypabyssal and subvolcanic (Cater, 1982). Intrusion depths are estimated to range from ∼25–30 km to as shallow as ∼7 km, with some individual plutons exposed for >8 km vertically (Miller et al., 2009).

The spectacularly outcropping Bergell intrusion in northeast Italy is reported to preserve a 10–12 km crustal transect, with no indication that deepest parts of the intrusion are exposed (Rosenberg et al., 1995; Berger et al., 1996; Samperton et al., 2013).

In northwest Italy, granite and granodiorite, exposed to paleodepths of ∼12 km, intrude the contemporaneous ignimbrite fill of the large Sesia caldera (Quick et al., 2009; Sbisà, 2010). This upper-crustal assemblage overlies the well-known lower-crustal mafic complex of the Ivrea-Verbano zone, both with similar Permian ages (ca. 290 Ma). Thus, a subcaldera magmatic plumbing system is exposed to ∼25 km depth, directly linking intrusion of mantle-derived basalt into the deep crust with large-scale silicic volcanism.

Crustal Sections through Magmatic Arcs

Tectonically tilted crustal sections through several continental-margin and oceanic arcs (Fig. 5) provide further documentation, on even larger scales, that subvolcanic Cordilleran batholithic bodies occupy much of the crust, with geochronologic implications for duration of pluton assembly and crystallization, total magma-body volume, and evidence for removal of their deep-crustal mafic roots (DeBari and Greene, 2011).

The Kohistan arc of northeast Pakistan contains an exceptional crustal section, from shallow volcanic and sedimentary deposits intruded by granites of the Kohistan batholith down to paleodepths of 30 km (Jagoutz and Schmidt, 2012, 2013). Mafic cumulates as residue from the shallower felsic magmatism are interpreted to dominate the lower crust down to the Moho at 50 km depth, but petrologic modeling suggests that the preserved cumulates are volumetrically inadequate by a factor of 2–3 to have generated the exposed upper-crustal batholith. Large-scale foundering of lower-crustal residue has been inferred to account for the volumetric discrepancies.

Structural and paleobarometric data indicate that exposed levels of the Sierra Nevada batholith deepen southward, from dominant granodiorite and quartz monzonite at upper-crustal levels preserving cogenetic Cretaceous volcanic rocks near Yosemite National Park (Fiske and Tobisch, 1994; Schweikert and Lahren, 1999), to tonalite–quartz diorite at paleodepths as great as 35–40 km in southern areas (Saleeby et al., 1990, 2003). Mafic residue (restites, cumulates) from batholith generation should have 1–2 times the volume of the granitic rocks, but seismic data, xenolith studies, and buoyancy considerations indicate that much of this residue has detached from beneath the batholith (Ducea, 2001; Saleeby et al., 2003; Jones et al., 2004; Zandt et al., 2004). In the Coastal plutonic complex of British Columbia and its southern extension into the North Cascades, Cretaceous to Paleogene (96–45 Ma) plutons of tonalite and lesser volumes of diorite record paleodepths of ∼7–35 km (Armstrong, 1988; Miller et al., 2009).

Exhumed and tilted sections in the Jurassic Talkeetna island arc expose upper-crustal volcanic sequences, 5–7 km thick, intruded by tonalites and quartz diorites and underlain by middle- and lower-crustal gabbros that are ∼15 km thick. The Talkeetna arc section overlies residual mantle that includes ultramafic cumulates (DeBari and Greene, 2011, and references). Similarly, in the Famatinian region of Argentina, tilted portions of an Ordovician continental-margin arc (485–465 Ma) expose an intact ∼25 km mid- to lower-crustal section of granodiorite-tonalite-gabbro (Ducea et al., 2010; Otamendi et al., 2012).

Additional perspectives on magma-assembly duration come from recent age determinations on SRMVF igneous rocks and other subvolcanic intrusions by 40Ar/39Ar and U-Pb methods that have much-improved analytical precision compared to earlier K/Ar and fission-track analyses. Before summarizing results for the SRMVF, these geochronologic techniques are reviewed briefly (building on excellent discussions by Miller et al., 2007; Simon et al., 2008; Schmitt, 2011), because applying such data to the interpretation of magma crystallization, supply rates, and duration of pluton assembly remains challenging. Various studies have attempted to test whether exposed subvolcanic plutons represent direct samples of a magma body that fed an eruption (Glazner et al., 2008; Tappa et al., 2011; Zimmerer and McIntosh, 2012a, 2012b), but no a priori reasons exist for such a magma body to undergo final crystallization and cooling without further modification, especially if associated with recurrent mafic recharge and multiple eruptions in a long-lived volcanic field. Many, probably most, subvolcanic magma bodies are likely to undergo compositional evolution over varied time intervals, with final crystallizing and cooling below solidus temperatures only as magma supply wanes. Alternatively, at some intrusions, decompression-induced volatile release may induce rapid crystallization in the nonerupted upper-crustal mush immediately after an eruption (Bachmann et al., 2012). Such a process could generate the common porphyritic textures in subvolcanic plutons. Perhaps more generally, waning of magma supply, which would disrupt quasi-equilibrium between long-term volatile release and magmatic recharge, would promote increased crystallization in subvolcanic intrusions at diverse time intervals after an eruption.

Single-crystal laser 40Ar/39Ar determinations can provide high-resolution measures of eruption age, especially for sanidine (with analytical uncertainties commonly now only 20–30 k.y. at 2σ for mid-Tertiary volcanic rocks), but they offer few insights concerning pre-eruption magma-chamber processes (Costa, 2008). Similarly, application of 40Ar/39Ar methods to minerals from granitic intrusions yields much evidence on rates of postcrystallization cooling but little information for duration of magma assembly or inception of crystallization (Memeti et al., 2010; Zimmerer and McIntosh, 2012a, 2012b). Despite the high analytical precision of 40Ar/39Ar methods, calibration of absolute age remains controversial (ages in this paper are normalized to Fish Canyon Tuff at 28.02 Ma, for consistency with recent SRMVF reports; Lipman and McIntosh, 2008; Lipman, 2012).

In contrast to the eruption ages for volcanic deposits, U-Pb determinations for granitic plutons, mainly on zircon, are likely to provide only partial records of potentially lengthy magmatic crystallization of this refractory mineral, and the timing of zircon growth relative to other mineral phases can vary greatly as a function of magma composition, liquidus-solidus temperature, and zircon saturation (Watson, 1996; Miller et al., 2007). Complex zoning of igneous zircons, some with internal resorption boundaries, likely records variable durations of fluctuating magma-chamber pressures and temperatures during assembly of subvolcanic magma bodies (Robinson and Miller, 1999; Miller et al., 2007; Schmitt, 2011; Erdmann et al., 2013). Complexities include open-system recharge by mantle-derived magma, mingling between contrasting compositional batches, and assimilation and mixing with partial melts of adjacent rocks, both earlier-solidified phases of the same broad magmatic system (antecrysts) and older crustal country rocks (xenocryts).

The analytical methods in most common use for zircon age determinations, thermal ionization mass spectrometry (TIMS), secondary ion mass spectrometry (SIMS), and laser-ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS), tend to yield partly complementary but partly conflicting results that have led to diverse interpretive approaches (see summary in Schmitt, 2011). Determining chronology of magmatic events by U-Pb dating of zircon seems somewhat analogous to dating the growth history of a forest by 14C measurements on wood fragments (resorbed zones in zircons are even mimicked by overgrown fire scars in old trees). (1) Dates for whole zircon crystals (by TIMS, with high precision) or dates on whole cross-sectional slabs of a tree trunk provide average ages for overall growth history of crystal or tree, respectively. (2) Dates on multiple fragments of a zircon crystal, or on broken chunks of a tree slab, provide a partial record of growth duration but inevitably underestimate the total time span. (3) Zircon core, rim, and surface ages (by SIMS or LA-ICP-MS, but with lower precision than TIMS), or ages for inner and outermost tree growth rings, provide the best age approximations for initial and last growth event. However, none of these methods provides reliable information on earliest growth history: for forests where long-dead trees have decomposed by biogenic processes, or for long-lived magma bodies that resorbed early-formed zircon when cycled through contrasting Zr-saturation environments.

Recent analytical innovations, including chemical-abrasion (CA) pretreatment (Mattinson, 2005), now provide remarkably precise ages by TIMS methods (some to 0.1% or less) for individual crystals of mid-Tertiary age, while reducing complications of Pb loss (Schaltegger et al., 2009; Schoene et al., 2012). High-precision CA-TIMS zircon ages for mid-Tertiary and older systems have documented lengthy crystallization histories for some ignimbrite magmas (Fig. 6A; Wotzlaw et al., 2013) and large plutons (Coleman et al., 2004; Memeti et al., 2010; Frazer et al., 2014). Zircons from some individual granitic samples yield prolonged semicontinuous age spectra (105–106 yr; Matzel et al., 2005, 2006b; Crowley et al., 2006; Schaltegger et al., 2009; Memeti et al., 2010), complicating interpretive distinction between autocrystic and antecrystic growth.

TIMS analyses are time-intensive, however, and many ages are reported as weighted means of relatively few individual zircon analyses, an approach that can obscure geologically informative information about duration of magma crystallization (von Quadt et al., 2011). Unsurprisingly, the range in apparent ages tends to increase with the number of individual crystals analyzed, especially for silicic (cool-wet) calc-alkaline magmas such as the SRMVF (Fig. 7A), which may never have been strongly zircon undersaturated. This positive correlation between age range and number of analyses suggests that weighted-mean TIMS ages are likely to underestimate duration of zircon crystallization, especially if the number of analyses is small or analyses are made on multiple zircon crystals.

Also, as averages for a crystal that may be zoned, TIMS ages will tend to underestimate duration of crystallization. Many recent TIMS studies interpret weighted-mean TIMS zircon dates as the crystallization age (e.g., Tappa et al., 2011; Mills and Coleman, 2013), while others use the younger zircon fractions as a record of late magmatic growth (e.g., Memeti et al., 2010; Schoene et al., 2012). Because much of total volume is in the outer zones of the crystal, the presence of an antecrystic core that grew a few million years earlier during evolution of a long-lived Tertiary magmatic system would be difficult to detect by even the highest-resolution TIMS analysis. For example, almost half the total volume is in the outer 20% of the radius in a simple spherical crystal, while less than 1% of the volume resides in the innermost 20%. Even a xenocrystic core of this size, derived from Proterozoic SRMVF basement (1800 Ma), would increase the whole-crystal age of a 35 Ma zircon growth by only 15 m.y.; a similar-size antecrystic core 3 m.y. older than the rim would increase the whole-crystal age by a nonresolvable 0.025 m.y. (at constant U content). As a further complication, metamict parts of zircons that are preferentially removed by CA pretreatment are likely to be high-U zones that grew from chemically evolved silicic phases, and this stage of growth history is thus likely to be poorly recorded by a bulk-crystal TIMS age. Many researchers in TIMS laboratories are currently exploring methods for improving resolution and precision by techniques such as combining ages with trace-element data on the same zircons (Schoene et al., 2012); dating multiple fragments from individual zircons (e.g., Samperton et al., 2013); or comparing matrix zircons with those enclosed in host crystals (Barboni and Schoene, 2014).

In contrast, much smaller portions of a crystal can be analyzed by SIMS methods, with analysis-spot diameters commonly ∼25 μm. Such spatial resolution has the potential to resolve core-to-rim differences in magmatic growth history (e.g., Brown and Fletcher, 1999), but typical precision for an individual analysis is at least an order of magnitude lower than for TIMS determinations on mid-Tertiary rocks. Despite such limitations, SIMS zircon studies have begun to provide important insights about duration of zircon growth. Core-rim zircon growth durations of as long as ∼5 × 105 m.y. have been documented for several Pleistocene volcanic units analyzed by in situ U-series and U-Pb methods (Fig. 6B), especially calc-alkaline arc rocks (Brown and Fletcher, 1999; Vazquez and Reid, 2004; Bachmann et al., 2007a; Reid, 2008; Simon et al., 2008; Claiborne et al., 2010; Schmitt et al., 2010).

Analyses by LA-ICP-MS also permit determinations on subareas of crystals with similar precision (Guillong et al., 2014), although spot size is somewhat larger, and excavation depth is greater than by SIMS methods. A counterbalancing advantage of LA-ICP-MS methods for some problems is rapid analysis, permitting generation of large data sets. Despite greater uncertainties for U-Pb SIMS and LA-ICP-MS determinations on older rocks in comparison to TIMS, antecrystic zircon cores as much as a million years or more older than the time of eruption, or final crystallization in intrusions, increasingly have been reported for Tertiary magmatic systems (some summarized here).

Zircon U/Pb ages thus can document times of zircon crystallization in the long-lived, incrementally assembled magmatic systems that are inferred here to generate vertically extensive plutons and batholiths beneath Cordilleran and Andean arcs, but contrary to some interpretations, such zircon ages provide only incomplete constraints on processes or overall duration of magma-body assembly. Early growth of a long-lived magmatic system may be poorly recorded due to periods of zircon dissolution, and the more-precise SIMS analyses of whole crystals may not resolve late stages. The construction of a batholith can take longer than recorded by zircon-crystallization ages, while the time interval for separation and shallow assembly of eruptible magma may be much shorter.

San Juan Locus

The San Juan volcanic region preserves the most numerous and diverse ignimbrites and source calderas in the composite SRMVF (Table 1); small granitic plutons are exposed at more than half the calderas, commonly intrusive into volcanic host rocks (Table 2). High-precision ages are available for all the ignimbrites and increasingly for the associated intrusions. Compositional, structural, and age relations for several of these systems provide special perspectives on processes and rates of magma supply and pluton assembly.

Fish Canyon Tuff, La Garita Caldera

The Fish Canyon Tuff, erupted at 28 Ma from the La Garita caldera, is among the best known ignimbrites in the SRMVF and worldwide because of its exceptional volume (>5000 km3) and widespread use of its crystal cargo as geochronologic reference standards. No associated granitic intrusions are exposed, but the tuff contains granodiorite fragments with zircon U/Pb ages indistinguishable from those of pumices (Bachmann et al., 2007c). Numerous recent age determinations on Fish Canyon sanidine and zircon provide critical insights about a complex magmatic history (see Phillips and Matchan, 2013, including references).

Fish Canyon sanidine ages, determined by 40Ar/39Ar, have proved highly consistent from crystal to crystal, despite complex resorption textures and compositional zoning in this phase (Bachmann et al., 2002) and common isotopic disequilibrium between crystal phases and the melt (Charlier et al., 2007). U-Pb determinations on Fish Canyon zircons have yielded ages modestly older than those for sanidine by 40Ar/39Ar methods, and recent high-precision studies have documented analytically significant durations of zircon crystallization (Fig. 6A).

Pre-eruption spans of zircon crystallization on the order of 450 k.y. in Fish Canyon magma have been documented by single-crystal CA-TIMS analyses for large zircon populations (19 crystals—Bachmann et al., 2007c; 58 crystals with particularly high precision—Wotzlaw et al., 2013). These age spans must be minimal, however, because even the most precise whole-crystal TIMS analyses provide no information on core-rim variation. The sizable whole-crystal age ranges require that the total duration of Fish Canyon zircon growth and magma assembly was significantly longer; final rim crystallization must have been later than the youngest whole-crystal dates (ca. 28. 2 Ma), and similarly earliest growth in zircon cores must have been earlier than the oldest dates (ca. 28.65 Ma). Consistent with such inferences, a detailed SIMS study of Fish Canyon zircons yielded a weighted-mean age of 28.0 Ma from unpolished zircon surfaces, while core ages from sectioned crystals were as much as 1.1 m.y. older (Coble, 2013). Although less precise than the TIMS results, the differences between the surface and core zircon ages suggest a substantially longer duration of zircon crystallization and magma-body assembly.

Neither zircon study detected xenocrystic cores with basement ages (although an earlier U-Pb study reported whole zircon crystals of Proterozoic age that were petrographically distinct in a bulk sample of Fish Canyon Tuff; Lanphere and Baadsgaard, 2001). Thus, the observed range in Fish Canyon zircon ages seems unlikely to be due to variable presence of small xenocrystic cores; the absence of ages that step back significantly toward that of the basement (1400–1800 Ma) suggests that assimilated crustal zircons were completely resorbed during early stages of magma-body assembly. The extended durations of zircon crystallization by both TIMS and SIMS methods, back in time at least to that of the preceding ignimbrite eruption from the same area (Masonic Park Tuff at ca. 28.7 Ma), is consistent with the inference of lengthy pre-eruption evolution, as previously developed from petrologic and isotopic evidence of disequilibria in Fish Canyon magma (Lipman et al., 1997; Bachmann et al., 2002; Bachmann and Bergantz, 2003; Charlier et al., 2007).

Bonanza Caldera

This source of the complexly zoned Bonanza Tuff, erupted at 33.15 Ma, preserves exceptionally diverse exposures: from topographic rim, thick intracaldera ignimbrite and overlying lava fill, to caldera floor and granitic intrusions in its steep resurgent dome (Lipman et al., 2013). The resurgent intrusions vary widely in composition and texture, from mafic granodiorite and andesite to silicic granite and aplite (56%–77% SiO2), a compositional range similar to that in the caldera-filling lavas. K-feldspar 40Ar/39Ar ages on lavas and resurgent intrusions are close to those from the tuff, to at most a few hundred thousand years younger, indicating emplacement and cooling of at least upper portions of the caldera-related magma body soon after the ignimbrite eruption.

Platoro Caldera Complex

After eruption of five large dacite ignimbrites from Platoro between 30.1 and 28.6 Ma, andesitic lavas filling the composite caldera were intruded by the 3 × 8 km Alamosa River stock. A 40Ar/39Ar biotite age of 27.98 ± 0.11 Ma from typical granodiorite (M. Zimmerer and W. McIntosh, 2013, written commun.) suggests final crystallization ∼0.5 m.y. younger than the last-erupted ignimbrite from Platoro and about concurrent with eruption of Fish Canyon Tuff from the central caldera complex.

Silverton Caldera

At the deeply eroded Silverton caldera, source of the 27.6 Ma Crystal Lake Tuff, the 4 × 8 km Sultan Mountain stock (among the largest exposed intrusions in the San Juan Mountains) intrudes older volcanic rocks along the southern caldera margin. A 40Ar/39Ar age of 26.6 ± 0.03 Ma on biotite (Bove et al., 2001) and a similar LA-ICP-MS age on zircon (Gonzales and Pecha, 2015) from granodiorite indicated that, in contrast to results from Bonanza, final crystallization of the Sultan Mountain stock was about a million years after the ignimbrite eruption and caldera formation at Silverton.

Other Caldera-Related Intrusions

Silicic plutons are also exposed at Uncompahgre, Lake City, South River, and San Luis calderas (Table 2). The Capitol City intrusion at Uncompahgre has a 40Ar/39Ar age about a million years younger than the associated ignimbrite (Bove et al., 2001), similar to the situation at Silverton. A resurgent intracaldera granodiorite in Alpine Gulch at Lake City has 40Ar/39Ar ages (biotite, K-feldspar) within analytical uncertainties of the associated ignimbrite (M. Zimmerer, 2014, written commun.). Intrusions at South River and San Luis were emplaced along caldera structures and are likely closely related in age but have not been dated directly.

Sawatch Range

Calderas aligned north-south along the crest of the Sawatch Range (Fig. 1) were sources of the earliest ignimbrite eruptions from the SRMVF. These include the 37 Ma Wall Mountain Tuff, erupted from an erosionally removed caldera above the Mount Princeton batholith, and the 34 Ma Badger Creek Tuff, erupted from the Mount Aetna caldera fragment just to the south (Table 1). In many respects, the 33 Ma Bonanza caldera within the NE San Juan Mountains is a transitional southern continuation of the Sawatch structural trend.

Recent detailed geochronologic studies of the Mount Princeton and Aetna intrusions by U-Pb and 40Ar/39Ar methods have provided impressively precise mineral ages that document complex relations between timing of the ignimbrite eruptions versus crystallization and cooling of subvolcanic intrusions (Fig. 7). Weighted-mean U-Pb zircon ages by CA-TIMS from three samples of Mount Princeton Quartz Monzonite (Fig. 7B) are 1–1.5 m.y. younger than the 37 Ma eruption age (40Ar/39Ar, sanidine) of the Wall Mountain Tuff (Mills and Coleman, 2013; Zimmerer and McIntosh, 2012a), leading these authors to infer that emplacement of the Mount Princeton Quartz Monzonite was unrelated to the ignimbrite magma. Differences of 0.1–0.3 m.y. in weighted-mean U-Pb ages among the three analyzed samples were interpreted to record discrete intrusion events and rapid cooling of separate small magma batches during incremental batholith assembly. Interpretations are complicated, however, by the small numbers of samples (3) and analyses per sample (4–7), some individual analyses on multiple grains, lack of dates from the earliest-crystallized margin of the batholith where in contact with Precambrian country rock (the fine-grained texturally heterogeneous Pomeroy phase of Toulmin and Hammarstrom, 1990), and a 700 k.y. spread among the analyses with a continuum of overlapping ages rather than discrete groupings for each sample (Mills and Coleman, 2013, their fig. 4). Alternatively, as discussed in a later section, these results may record late zircon crystallization in a vertically extensive long-lived magma body from which the Wall Mountain Tuff had erupted. Ages on hornblende and biotite from this pluton document relatively rapid cooling after zircon growth (Zimmerer and McIntosh, 2012a) but offer no direct information on timing of magma assembly.

In contrast to results for the deeply eroded Princeton intrusion, two of three granitic samples from the Aetna caldera yielded CA-TIMS zircon ages indistinguishable from that of the associated Badger Creek Tuff at 34.5 Ma (Mills and Coleman, 2013). Zircons from a third sample have a distinctly older weighted-mean age (34.95 ± 0.04 Ma), even though this body as presently mapped intrudes the intracaldera Badger Creek Tuff. The dated sample has been suggested to be from an unmapped older intrusion (Coleman et al., 2013), but this interpretation would require that the older body was erosionally exposed extraordinarily rapidly, prior to deposition of the Badger Creek Tuff. Alternatively, the 34.95 Ma weighted-mean age could reflect antecrystic zircon growth not detectable by whole-crystal TIMS analyses, perhaps extending back in time to assembly of the Princeton Quartz Monzonite.

In addition to the 40Ar/39Ar age and CA-TIMS zircon determinations for Princeton and Aetna intrusions, a few of the same samples were dated by LA-ICP-MS (Zimmerer and McIntosh, 2012a). Analysis of 25–35 zircons per sample yielded weighted-mean dates similar to those by TIMS methods, but with a much broader spectrum of individual ages (Fig. 7B). Ages statistically younger than the weighted mean were inferred by Zimmerer and McIntosh to reflect Pb loss, a potential complexity also noticed in other mid-Tertiary granitic rocks (DuBray et al., 2011; Colgan et al., 2012; von Quadt et al., 2014). Ages that were statistically older than the weighted mean, some as old as 38 Ma and similar to the eruption age of the Wall Mountain Tuff, were inferred to record early zircon growth, well before final solidification of this intrusion. Such interpretations were tempered, however, by the relatively large uncertainties of the LA-ICP-MS technique.

Questa-Latir Area

Recent U-Pb and 40Ar/39Ar age determinations from the Latir volcanic field, the Questa caldera source of the 25.5 Ma Amalia Tuff, and associated intrusions have yielded notably precise and coherent results (Tappa et al., 2011; Zimmerer and McIntosh, 2012b; Rosera et al., 2013), with broad similarities to those from the Sawatch Range. This younger ignimbrite-caldera system preserves a more complete proximal volcanic sequence and compositionally diverse associated granitic intrusions. The intrusions are exposed progressively deeper southward from the caldera, but gravity data show that the isolated exposures are upper parts of a large composite batholith (Cordell et al., 1985; Lipman, 1988).

The shallowest intracaldera resurgent intrusion (peralkaline granite of Virgin Canyon) has a composition and age similar to the Amalia Tuff, and the slightly deeper metaluminous Rito del Medio and Canada Pinabete granitic plutons yield CA-TIMS crystallization ages for zircon that are at most a few hundred thousand years younger (Tappa et al., 2011). Two samples from the deeper and less-evolved granodiorite of Cabresto Lake have slightly younger crystallization ages (25.1–25.0 Ma), and mineralized granitic intrusions along the southern ring-fault zone are variably still younger, 25.2–24.5 Ma (Rosera et al., 2013). South of Questa caldera, four samples from the large Rio Hondo pluton, which has its roof zone entirely within Precambrian basement, yielded weighted-mean TIMS ages from 23.0 to 22.6 Ma, but the individual ages define a continuum without analytically significant gaps between samples (Fig. 8A, inset; Tappa et al., 2011).

These authors interpreted the progression of zircon ages, which get younger southward from the caldera area, as recording incremental top-down emplacement and rapid crystallization of separate small magma batches, analogous to the interpretations of TIMS zircon results from the Sawatch Range by Mills and Coleman (2013). Alternatively, could at least parts of the 2–3 m.y. crystallization span of the Questa-Latir intrusive suite record protracted assembly of an intermittently reactivated crustal-scale magmatic system?

Other Cordilleran-Arc Systems

Recent studies of volcanic areas and associated upper-crustal batholiths elsewhere in the Cordillera provide additional perspectives that augment data available from the SRMVF. Results from a few especially pertinent areas are summarized briefly here.

Organ Caldera and Batholith, New Mexico

The Organ Mountains, farther south in the Rocky Mountains of New Mexico, expose a composite batholith, from surface levels to >6 km depth, that intrudes a thick intracaldera ignimbrite (Seager and McCurry, 1988). Combined 40Ar/39Ar sanidine and sparse U-Pb zircon data show that ignimbrite and intrusions are closely similar in composition and age, at 36–35 Ma (Zimmerer and McIntosh, 2013; Rioux et al., 2010). These relations are comparable to those at the Bonanza and Aetna calderas but contrast with the more deeply eroded Mount Princeton batholith.

Basin and Range, Nevada and Utah

At the Caetano caldera, Nevada, rotational basin-range faulting exposes exceptionally thick (∼5 km) intracaldera rhyolitic ignimbrite, intruded by granite that has an identical 40Ar/39Ar age (34.0 ± 0.5 Ma) within analytical uncertainties (John et al., 2008; Henry and John, 2013). Zircon SIMS determinations on both ignimbrite and granite also have peak ages at 34 Ma, but include younger and older outlier values (John et al., 2009). Pretreatment of zircons from Caetano ignimbrite and granite by chemical abrasion reduced numbers of anomalously young ages while preserving and augmenting proportions of statistically significant determinations that indicate presence of antecrystic zircon as much as 2 m.y. older than eruption of the associated ignimbrite (Colgan et al., 2012). Mid-Tertiary ignimbrites and shallow caldera-related plutons have yielded similar age concordances in the Stillwater and Clan Alpine Mountains (John et al., 2014).

At the mineralized porphyry in Bingham Canyon, Utah, the oldest and youngest intrusive phases have yielded zircon ages within the range 38.1–37.8 Ma, but in each phase, several concordant zircons are significantly older, ranging back to at least as far as 38.5 Ma and possibly to ca. 40.5 Ma (von Quadt et al., 2011). These variations were interpreted as a minimum lifetime of the magma reservoir of at least 0.7 m.y., possibly more than 2 m.y. The Caetano and Bingham results are thus comparable to the documented antecrysts in the Fish Canyon Tuff (Wotzlaw et al., 2013; Coble, 2013) and to less-robust data suggesting the presence of antecrystic zircons at Mount Princeton (Fig. 7B; Zimmerer and McIntosh, 2012a).

Sierra Madre Occidental, Mexico

Several large-volume rhyolitic ignimbrites from this vast Cordilleran-arc province have yielded U-Pb zircon populations 1–4 m.y. older than the eruption age as determined by K/Ar and 40Ar/39Ar; the range of ages was interpreted as indicating derivation of much of the zircon by remobilization of partially molten to solidified rocks formed during preceding phases of Sierra Madre volcanism (Bryan et al., 2008). In the volcanic province of Sierra Madre del Sur, field relations and LA-ICP-MS zircon ages from the Tilzpotla caldera and batholith suggest prolonged semicontinuous assembly of a volcano-plutonic system starting ca. 39.5 Ma, which climaxed with eruption of the Tilzpotla ignimbrite and associated caldera at 34.3 Ma. Cores of individual zircons have ages as much as 2 m.y. older than rims or K-Ar ages for the same unit (Martiny et al., 2013, their table 3).

Altiplano-Puna Volcanic Complex, Andes

The well-documented APVC of the central Andes (de Silva, 1989; Schmitt et al., 2002; de Silva et al., 2006; Salisbury et al., 2011), other than its younger age (10–1 Ma), is closely comparable to the SRMVF (table 4 inLipman and McIntosh, 2008). At Cerro Galan caldera, SIMS zircon ages from a sequence of nine large silicic ignimbrites erupted between 2.0 and 5.6 Ma provide direct evidence of prolonged crystallization (Folkes et al., 2011). Interiors of zircons commonly crystallized up to several hundred thousand years prior to eruption, and zircons from many ignimbrites contain antecrystic cores from a previous cycle in the Cerro Galan sequence. At the long-lived, lava-dominated Aucanquilcha volcanic cluster, semicontinuous zircon age spectra from individual samples extend to as much as 2 m.y. older than eruption ages, durations consistent with widespread dissolution of antecrystic zircon during the thermal peak of magmatism (Walker et al., 2010).

Elkhorn Mountain Volcanics and Boulder Batholith

The 50 × 100 km Butte Quartz Monzonite and associated smaller plutons of the Boulder batholith were intruded to shallow crustal levels, roofed largely by ignimbrites of the closely related Elkhorn Mountain volcanics (Robinson et al., 1968; Tilling, 1974; Hamilton and Myers, 1974). Recent SIMS determinations document zircon crystallization in plutons from 81 to 73 Ma, an 8 m.y. span that is comparable to the duration of ignimbrite eruptions in continental-margin arcs such as the SRMVF and APVC. Internal resorption boundaries and core to crystal-surface differences of as much as several million years within individual zircons have been interpreted to record complex processes, including multiple episodes of crystallization and magma-reservoir replenishment that spanned several million years during zircon growth (Aleinikoff et al., 2012).

Adamello Batholith, Northern Italy

Recent CA-TIMS zircon studies at the 43 to 33 Ma Adamello batholith have provided perhaps the most precise data thus far for any comparably complex granitic system (Schaltegger et al., 2009; Schoene et al., 2012). Zircon ages from the composite Re di Castello pluton at the southern end of the batholith have been interpreted to indicate 1.5 m.y. of crystallization, with zircons from individual samples spanning as much as 700 k.y. (Schaltegger et al., 2009). Because the analyses are from whole zircons, the measured durations of growth likely are minima for assembly and crystallization.

Sierra Nevada Batholith

Documentation of an ∼8 m.y. span of zircon ages (94–86 Ma) from granitic units of the concentrically zoned Tuolumne igneous suite (Fleck and Kistler, 1994; Coleman et al., 2004), followed by similar results for other Sierran intrusions (Davis et al., 2012; Frazer et al., 2014), has stimulated much controversy concerning assembly duration and crystallization history of Cordilleran plutons. The multimillion-year durations of zircon crystallization in these plutons are impressively longer those obtained by similar methods for shallower subvolcanic intrusions such as those associated with the SRMVF. The large spans of Sierran zircon ages have been interpreted as requiring incremental intrusion in small sheet-like batches, followed by rapid solidification that precluded existence of eruptible magma comparable in volume to large ignimbrites. Processes of pluton-batholith construction were thus interpreted as disconnected from generation of magma bodies capable of large-volume explosive volcanism (Glazner et al., 2004, 2008; Bartley et al., 2005; Mills and Coleman, 2013; Frazer et al., 2014).

Other results, especially for CA-pretreated single crystals, have documented additional complexities in the zircon geochronology of Sierran plutons (Matzel et al., 2006b; Crowley et al., 2006; Memeti et al., 2010; Frazer et al., 2014), broadly similar to those just summarized for ignimbrites such as the Fish Canyon Tuff (Fig. 8B). Multiple crystals from some individual samples define a continuum of ages spanning as much as several million years (Fig. 8C). As evaluated here, these ages provide insights into the duration of late crystallization as pluton construction waned, but they may preserve only an incomplete record of earlier events during potentially prolonged assembly intervals. Such relations also bring into question whether reliable information concerning duration of magma assembly and pluton construction can be obtained from weighted-mean ages from multiple analyses of whole-zircon crystals (Miller et al., 2007; Schmitt, 2011; de Silva and Gregg, 2014).

Because geochemical evidence is compelling that generation of SRMVF and other Cordillera-arc magmas involved voluminous melting and assimilation of crustal basement, providing as much as 50% of erupted magma (e.g.,, DePaolo, 1981; DePaolo et al., 1992; Johnson, 1991; Perry et al., 1993; Ducea and Barton, 2007; Farmer et al., 2008; Kay et al., 2010), the presence or absence of basement-derived zircon provides information on processes of magma evolution and pluton assembly. The presence of even small relict zircon cores from Proterozoic rocks (1400–1800 Ma) beneath the SRMVF should be recognizable; a relict core only 1% by volume would increase apparent age of a mid-Tertiary zircon by ∼15 m.y.

Survival of basement xenocrysts could imply either (1) a low-temperature history of magma generation and pluton assembly, which appears inconsistent with compositions of many SRMVF intrusions (relatively mafic granodiorite), or, more likely, (2) rapid magma generation, emplacement, and crystallization. Alternatively, sparseness or absence of zircon xenocrysts in SRMVF magmas would imply initial melting/assimilation of basement rocks in a Zr-undersaturated environment and/or dissolution during episodes of later magmatic recharge at temperatures above zircon saturation. The absence of xenocrysts is unlikely to have resulted from separation of crustal melts from a restite that retained zircon crystals, because andesitic to dacitic SRMVF magmas have ordinary to relatively high Zr contents for calc-alkaline systems (mostly in range 175–250 ppm; e.g., CD-ROM tables inLipman, 2006, 2012).

Evaluation of xenocryst distribution as functions of diverse variables (pluton composition, age, size, relation to ignimbrite eruption) is limited by sparse data for SRMVF plutons, as well as uncertainties resulting from diverse sample processing. Typically, only small numbers of zircons have been analyzed in TIMS studies, and many of the larger SIMS and LA-ICP-MS data sets are dominated by rim and surface analyses in effort to date final crystallization. Also, some studies may have preferentially analyzed zircons characterized by simple morphology and texture, while others have analyzed more diverse crystal suites.

Despite such limitations, existing data suggest that basement zircon xenocrysts are rare in SRMVF intrusions that are closely related to ignimbrite calderas within areas of the geophysically defined subvolcanic batholith, while at least some outlying Tertiary intrusions, especially those of relatively small size, have more diverse zircon populations (Table 4; Fig. 9). Particularly informative are relatively large LA-ICP-MS data sets that document at most a few percent of basement-derived zircons in caldera-related intrusions: Princeton batholith (3 of 238 analyses: 8 samples), Aetna (2 of 65 analyses: 2 samples), and Organ (1 of 185 analyses: 7 samples), as reported by Zimmerer and McIntosh (2012a, 2013). In contrast, some outlying western San Juan intrusions peripheral to the geophysically defined batholith contain abundant basement-age zircons; e.g., more than 50% of analyses from the Ophir and Wilson stocks (Table 4; Gonzales and Pecha, 2015).

Zircon age populations are likely to be more complex and difficult to interpret in ignimbrites, because eruption and transport processes may incorporate xenocrysts from the vent and/or ground surface. For example, 10 zircon determinations for an APVC ignimbrite pumice yielded a tightly clustered array defining an age of 4.65 ± 0.13 Ma, in agreement with the K/Ar biotite age for the same sample, but 13 zircons from a bulk sample of this ignimbrite yielded a spectrum of older APVC ages (9–13 Ma), as well as some Paleozoic basement ages (Schmitt et al., 2002). Nevertheless, several detailed zircon studies of the Fish Canyon Tuff detected no surviving Precambrian component in this large SRMVF ignimbrite (Bachmann et al., 2007c; Wotzlaw et al., 2013; Coble, 2013). For the 34 Ma Badger Creek Tuff, erupted from Mount Aetna caldera, only one of 31 zircons in a bulk sample analyzed by LA-ICP-M has a Proterozoic age (Zimmerer and McIntosh, 2012a). Both these crystal-rich dacitic ignimbrites contain complexly resorbed crystal cargos, indicative of recharge and mixing with mafic magmas, yet, like associated SRMVF intrusions, they retain little record of crustal melting and assimilation during magma generation. Precambrian zircon ages were modestly more abundant in a bulk sample of Wall Mountain Tuff (7 of 40 LA-ICP-MS spot analyses; Zimmerer and McIntosh, 2012a), perhaps representing crystals from vent walls and the ground surface during eruption and emplacement of this first large ignimbrite in the SRMVF. Contrary to the inference that zircon dissolution should be slower than magma-body assembly in silicic magmas (Frazer et al., 2014), basement zircon xenocrysts are sparse to absent in many other major ignimbrites that have been well characterized: the 161 ka Kos Plateau Tuff in Greece (Bachmann et al., 2007a; Guillong et al., 2014), the mid-Tertiary ignimbrites of the Sierra Madre Occidental (Bryan et al., 2008; Martiny et al., 2013), the Late Tertiary La Pacana and Cerro Galan ignimbrites of the APVC (Schmitt et al., 2002; Folkes et al., 2011), the 35 Ma Caetano Tuff and associated granitic intrusions in Nevada (John et al., 2009; Colgan et al., 2012), the 18.8 Ma Peach Springs Tuff in Arizona (McDowell et al., 2014), and the Oligocene Black Mountain, Chiquito Peak, Carpenter Ridge, and Nelson Mountain Tuffs of the SRMVF (M. Zimmerer and M. Verdon, 2014, written commun.).

Thus, the apparent paucity of xenocrystic basement zircons both in large silicic ignimbrites and in caldera-related intrusions (Table 4) favors interpretations of prolonged magma-body assembly accompanied by intermittent crystal dissolution, involving temperature cycling above/below zircon saturation during mafic-recharge events, as also proposed in some reports cited earlier. Critical variables in determining conditions under which preexisting zircon crystals could be resorbed during recharge events (Watson, 1996; Frazer et al., 2014) likely include melt composition, pressure, temperature, and volatile content, duration of such Zr-undersaturated intervals, and stage during construction of the magma body. A related complexity is that a growing system may become increasingly armored to contamination by old basement as earlier-formed intrusions comprise the host rock for later magmatic increments, but even the earliest-erupted ignimbrites in the SRMVF and elsewhere in the Cordillera seem generally to lack zircon xenocrysts.

The contrasts in xenocryst populations between small distal versus caldera-related intrusions (Fig. 9) also suggest that plutons and batholiths constructed by multiple small intrusive batches that crystallized rapidly, as proposed for SRMVF intrusions (Tappa et al., 2011; Mills and Coleman, 2013) and for Sierra Nevada intrusive suites (Glazner et al., 2004; Davis et al., 2012; Frazer et al., 2014), should contain bimodal zircon populations. At least some samples would be predicted to preserve abundant basement xenocrysts along with a tight cluster of zircon ages that record autocryst growth during rapid cooling of the intrusive batch, rather than the overlapping broad spectra of ages (105–106 yr) commonly observed by TIMS analyses for subcaldera and deeper plutons (Figs. 7–8).

Further insights concerning magma assembly and crystallization come from age and depth relations between caldera-related intrusions and associated ignimbrites (Fig. 10). Data from the SRMVF and elsewhere suggest that shallow caldera plutons, especially ones that resurgently intrude caldera-filling tuff, tend to have crystallization ages similar to those of the associated ignimbrite (Table 5; Fig. 10A). In contrast, intrusions at deeper levels, beneath or adjacent to an ignimbrite caldera, tend to be variably younger. For example, the diverse intrusions associated with the Questa caldera decrease in age from 25.5 to ca. 23 Ma (Tappa et al., 2011; Rosera et al., 2013) as exposure depth deepens southward (Lipman, 1988). Deeper intrusions such as Rio Hondo, south of the caldera, have even younger crystallization ages (23.1–22.6 Ma; Tappa et al., 2011), raising interpretive questions about genetic relation to the ignimbrite magma. A schematic interpretation, based on zircon-crystallization ages in subcaldera SRMVF intrusions (Fig. 10B), infers large variations in upper portions of vertically extensive subcaldera plutons but overlapping ages at greater depths.

The age-depth correlations are variable due in part to limits of analytical resolution, but they also reflect geologic variability. Some intrusion ages plotted in Figure 10A have large uncertainties, especially LA-ICP-MS results, and some U-Pb crystallization ages may be anomalously old because of presence of antecrysts. Continued intrusive activity preferentially along caldera structures would be expected at long-lived magmatic systems for varied intervals after an ignimbrite eruption. Magma-body assembly that leads to a large ignimbrite eruption would involve major disruption of the upper crust, generating zones of structural weakness that could facilitate continued emplacement of younger intrusions. In addition, a recurrently replenished long-lived magmatic system need not wane at the time of any single large ignimbrite eruption. For example, in the SRMVF, multiple large ignimbrites were erupted during prolonged spans from within the same site, such as the five ignimbrites from Platoro caldera complex at 30.1–28.6 Ma, or the nine tuffs from the central cluster at 28.7–26.9 Ma (Table 1). Such recurrent large eruptions, within time spans that are relatively brief in comparison to durations of zircon crystallization documented for some plutons (Figs. 7–8), provide direct evidence for long-lived recurrently reactivated magma bodies.

In contrast, some relatively small upper-crustal intrusions record rapid cooling and crystallization over sizable depth ranges. Near-surface and deep samples from a 4-km-deep drill hole into the geologically young Eldzhurtinsky Granite (Tirniauz, Russia), which may be associated with an ignimbrite eruption (Gurbanov et al., 2004), have yielded analytically indistinguishable weighted-mean SIMS-zircon ages of 2.04 ± 0.03 Ma; slightly greater dispersion of ages in the deeper sample likely reflects slower cooling (Grun et al., 1999). Biotite 40Ar/39Ar ages from the drill hole decrease with depth from 1.9 to 1.5 Ma and are also interpreted as recording cooling history (Gazis et al., 1995). Multiple sheets at the shallow Torres del Paine laccolith of Miocene age (12.5 Ma) in southern Chile are interpreted from TIMS zircon ages as recording assembly within 90 ± 40 k.y. (Michel et al., 2008). Similar processes and durations likely characterize other relatively small tabular intrusions.

In addition to age-depth variations for large caldera-related intrusions, deeper intrusions appear to have longer time spans of zircon crystallization, based on CA-TIMS data among crystals from individual samples. In the SRMVF (Fig. 8), zircon age spans from single samples of relatively shallow caldera-related plutons are typically in the range 0.2 m.y. or less at Questa (Tappa et al., 2011; Rosera et al., 2013) and at Princeton-Aetna (Mills and Coleman, 2013). In contrast, individual samples from granitic plutons at deeper intrusion levels have commonly yielded longer spans: 0.7 m.y. at Adamello (Schaltegger et al., 2009), as much as 1.2 m.y. in the North Cascades (Matzel et al., 2006a), and 0.5–2.5 m.y. at the Tuolumne complex (Memeti et al., 2010). At the Bergell intrusion (Italy), age spans within individual samples are about as great as the span of weighted-mean ages from the entire pluton (Samperton et al., 2013). The longer spans in some deeper intrusions might be due to presence of antecrysts unrelated to the active magma-assembly event (Matzel et al., 2005; Miller et al., 2007; Memeti et al., 2010), but the common absence of analytically detectable time gaps among such zircon age populations (Fig. 8C, inset) permits interpretation that these semicontinuous age spectra provide a partial record of earlier assembly and crystallization during long-lived construction in deeper parts of a mushy pluton (Fig. 10B), involving sustained magma supply, slow cooling, and intermittent zircon crystallization.

Perhaps distinction should be made between antecryst populations for which ages suggest a continuum of crystallization (e.g., Fish Canyon Tuff—Fig. 6A; Kos Plateau Tuff—Bachmann et al., 2007a; Guillong et al., 2014; Aucanquilcha—Walker et al., 2010; and granitic plutons noted earlier herein) versus those where analytically significant gaps exist within crystallization-age spectra (e.g., Taupo—Brown and Fletcher, 1999; Charlier et al., 2005; Crater Lake—Bacon and Lowenstern, 2005; Mount St. Helens—Claiborne et al., 2010; Tarawerea—Storm et al., 2012). Such apparent gaps in crystallization history need not imply subsolidus cooling. They may result from periods of slow growth (e.g., Watson, 1996), armoring within other mineral phases that interrupted growth, or prolonged storage of melt and crystals that interacted intermittently with hotter mafic influxes before erupting (e.g., Miller et al., 2007; Schmitt et al., 2010; de Silva and Gregg, 2014). Gaps in antecryst populations seem more likely to survive and record significant magmatic events at central volcanoes that erupt relatively small volumes than during prolonged crystallization in larger long-lived systems that lead to ignimbrite supereruptions. In such large systems, zircon ages are inferred to preferentially record younger episodes, interrupted by periods of thermal and compositional rejuvenation, and provide an incomplete record of early magma assembly.

While most of the dated shallow caldera-related intrusions in the San Juan region are similar in age to the associated ignimbrite (Table 2), some intrusions near the margins of the geophysically defined batholith and most of the more-distal ones have younger ages (ca. 26.5–5 Ma). Major concentrations of such batholith-margin intrusions include 23 to 20 Ma dikes and subvolcanic plugs of dacite and rhyolite along the north side of the Platoro caldera complex (Lipman, 1975; Lipman, W. McIntosh, and M. Zimmerer, 2013, written commun.) and small intrusions of similar composition within and along the west margin of the Silverton caldera that have yielded ages of 26–10 Ma (Lipman et al., 1976; Bove et al., 2001). Distal sills and laccoliths of monzonite to granite and porphyritic rhyolite, including bodies near Ophir and Rico west and southwest of Silverton, have ages from 26 to 5 Ma (Naeser et al., 1980; Cunningham et al., 1994; Gonzales and Pecha, 2015).

These widely distributed post-ignimbrite intrusions of Miocene age record the waning of San Juan magmatism (Fig. 11). Concurrent with mafic lavas of the Hinsdale Formation (also ca. 26–5 Ma), the late intrusions include more-evolved silicic compositions than in the Oligocene caldera-related magmatism; they have long been interpreted as components of bimodal magmatism associated with regional extension and opening of the Rio Grande rift zone (Lipman et al., 1970; Lipman and Mehnert, 1975; Thompson et al., 1991).

In contrast, no silicic intrusions or lavas of Miocene age have been identified centrally within the geophysically defined batholith, and timing of the shallow batholith-related magmatism appears broadly antithetic to that of the distal intrusions (Fig. 11). Only volumetrically minor mafic Hinsdale lavas (silicic basalt and basaltic andesite), erupted from dikes or pipes, seemingly were able to penetrate central parts of the Oligocene batholith (Lipman and Mehnert, 1975). Major additions to the batholith during the Miocene, after waning of the ignimbrite eruptions, seem improbable without surface volcanism. Rather, the regionally contrasting space-time-compositional relations suggest that only small volumes of silicic magma were generated during the Miocene. These magmas were unable to penetrate central parts of the batholith, perhaps because still-warm subsolidus granitic rocks that occupied much of the subvolcanic crust impeded the rise of intrusions more silicic than basalt.

Especially challenging goals for volcano-plutonic studies have been determination of the supply rates and assembly durations of subvolcanic magma bodies (Annen, 2009; Paterson et al., 2011; Tappa et al., 2011; Gelman et al., 2013). Some recent geochronologic and modeling studies have tried to use variations in zircon-crystallization ages among surface samples of a pluton to compute duration of magma-body assembly. Combined with estimates of exposed intrusion volume, such age ranges have been interpreted to indicate incremental tabular assembly of upper-crustal granites and “pluton-fill” rates (0.002 km3/yr or less) that would be too low to accumulate magma capable of supplying large ignimbrite eruptions (e.g., Glazner et al., 2004; Bartley et al., 2005; Annen, 2009; Tappa et al., 2011; Davis et al., 2012; Mills and Coleman, 2013; Schöpa and Annen, 2013; Frazer et al., 2014). As discussed previously, however, the sample-to-sample variations in weighted-mean ages are open to alternative interpretations, and none of these studies documented systematic age-depth progressions within a pluton. Additionally, low pluton-filling rates, obtained from the elevation range among surface exposures (e.g., ∼1 m.y./km at Mount Princeton and Rio Hondo plutons in the SRMVF; as much as ∼4–7 m.y./km for Sierran intrusions such as Half Dome or Mount Givens; Mills and Coleman, 2013; Frazer et al., 2014), would imply improbably lengthy assembly durations (∼100 m.y.!) for vertically extensive plutons (20 km or more) as inferred here for SRMVF intrusions and those in tilted crustal sections (Fig. 5).

In contrast to the large differences in U-Pb zircon ages within many samples (Fig. 8) and in relation to depth below associated volcanic rocks (Fig. 10A), some plutons exposed over multi-kilometer ranges have little detected variation in zircon age with depth: The top and bottom of the 4 km drill-hole section into the Eldzhurtinsky granite are indistinguishable at 2.04 ± 0.03 Ma (Grun et al., 1999); the bulk of the 10-km-thick Spirit Mountain pluton is within analytical uncertainty at 16.0–15.7 Ma (Walker et al., 2006; Miller et al., 2011); ages from the roof zone and deepest exposed level in the ∼12 km tilted section through the Jurassic Wooley Creek pluton differ by only ∼0.5 m.y. (Coint et al., 2013); dates from the 12 to 15 km depth of exposures through the Bergell intrusion vary by ∼2 m.y. (32–30 Ma, only modestly greater than age ranges within single samples; Samperton et al., 2013). Such limited variations in age with depth suggest widespread crystallization deep in plutons mainly late during assembly as magma supply waned (Fig. 10B). The “fill-rate” approach, based on averaged whole-crystal zircon ages, is likely to mainly date late crystallization, incompletely recording early stages of long-duration pluton construction. Also, interpretations made solely from surface exposures are likely to underestimate the total magmatic volume and thermal budget in vertically extensive magma bodies generated by voluminous mantle input.

Assembly of the SRMVF Batholith

Alternatively, duration of batholith assembly and time-space fluctuations in focused magma supply may be inferred from eruptive history, while overall volume of the magmatic system can be estimated from geochemical and isotopic data in conjunction with geophysical constraints on batholith area and vertical extent. Using such an approach, age and volume estimates from the SRMVF may provide useful broad insights concerning long-term magma-supply rates and pluton-assembly processes (Table 6; Fig. 12A), even though such rates and processes were likely modulated by shorter-term fluctuations and areal magmatic focusing. Varied time and volume approximations are explored, from the scale of individual ignimbrite-caldera cycles (Creede, La Garita), to the composite San Juan batholith, and the entire magmatic system of the SRMVF. Magma volumes and supply rates are estimated separately for the volcanic rocks, the crustal granitoid batholith, and the overall magmatic system, including mafic input from mantle sources (Table 6).

Alternate assembly durations for construction of the San Juan batholith and associated volcanic ejecta (∼400–500 × 103 km3; Table 3) could be 2–8 m.y., depending whether peak magma supply was focused during the caldera-forming eruptions (28.7–26.8 Ma) or spread more uniformly over the entire time of volcanism (34–26 Ma). For the entire span, resulting total magma-supply rate for the batholith is 0.04–0.06 km3/yr (Table 6), at the high end of supply rates for crustal systems (Crisp, 1984; White et al., 2006). If the bulk of the batholith was assembled during the ignimbrite eruptions, (all San Juan calderas except the early Platoro system lie above the gravity-defined batholith), the total magma supply could have been much higher during that interval (0.18 km3/yr; Table 6). For magma-supply focused at areas of individual caldera cycles, such as La Garita or Creede during the interval since the prior ignimbrite eruption, estimated supply rates are similarly high.

An analogous calculation for construction of the entire SRMVF and associated intrusions (38–26 Ma, total estimated magma volume ∼620,000 km3) still yields an average total magma-supply of 0.05 km3/yr, but across a vast area. However, the widely distributed intermediate-composition stratovolcanoes that preceded the ignimbrite eruptions were fed by small central conduits and radiating dike systems, many of which lie beyond the main gravity low, lack discernible geophysical expression, and seem less likely to have been associated with high rates of pluton construction.

These San Juan and SRMVF magma-supply rates are 10–20 times the threshold values required to grow and maintain crustal magma reservoirs sufficiently large to feed supereruptions, as estimated from caldera-scaled thermal models (∼0.01 km3/yr—Annen, 2009; or as low as ∼0.005 km3/yr allowing for nonlinear crystallization and temperature-dependent wall-rock conductivity—Gelman et al., 2013). The SRMVF values are also 30–100 times greater than representative pluton-fill rates (Table 7) estimated from age variations among surface samples of other Cordilleran plutons (Glazner et al., 2004; Davis et al., 2012; Mills and Coleman, 2013; Frazer et al., 2014). Magma supply for such plutons would become an order of magnitude or more higher, however, if modeled as vertically extensive (∼20 km thick) and involving subequal volumes of mafic residue (Table 7; Fig. 12B). As a further complexity, the thermal models published to date do not consider the potentially substantial effects from voluminous mafic mantle input into the crust.

The San Juan and SRMVF supply rates are for the batholithic-scale area, however, which are much larger than plausible sizes of individual caldera-related plutons such as the 10 km pluton radius explored by the thermal models of Annen and Gelman. Average supply rates per unit area of the entire SRMVF would be lower, ∼0.005 km3/yr for an area comparable to the pluton geometry modeled by Annen (2009) or Gelman et al. (2013), i.e., at the low end of supply values to support generation of a large ignimbrite eruption. No model of constant magma supply seems plausible or necessary throughout the assembly time and across the overall area of a large composite batholith, however, and sizable fluctuations in magma supply rate, areal focusing of pluton assembly, and wall-rock structural response seem inevitable (Paterson et al., 2011). In-progress thermal modeling by S. Gelman and O. Bachmann is showing that the estimated magma flux for the San Juan batholith, sustained for several million years, could have generated ignimbrite-size volumes of eruptible magma from chambers with horizontal dimensions on the scale of SRMVF calderas. Such large reservoirs buffer temperature and generate ductile wall-rock halos, promoting survival of long-lived chambers that undergo complex crystallization at near-zircon-saturation temperatures and provide recurrent opportunities to recycle, resorb, and reset geochronologic clocks (Bachmann et al., 2007b; Lipman, 2007; Gelman et al., 2013; Gregg et al., 2012).

Volumetric and compositional variations in magma flux within volcanic systems and associated intrusions have long been recognized as major controls on eruptive processes and pluton assembly (e.g., Smith, 1979; Hildreth, 1981; de Silva, 1989). Such fluctuation and focusing would be consistent with growth of the San Juan magmatic locus and the space-time variations in ignimbrite eruptions (Fig. 13): Sizable areas centrally within the geophysical batholith lack calderas, and multimillion-year pauses between ignimbrite eruptions alternate with successive eruptions that are too brief to resolve (e.g., three large ignimbrites in <0.1 m.y. from the San Luis caldera complex; Lipman and McIntosh, 2008). While the record of magmatic focusing in space and time seems clear in the volcanic record and probably in associated shallow intrusions, long-term deep regional flux may be more nearly constant, as suggested by the lengthy crystallization recorded by deeper intrusions in the Sierra Nevada (Coleman et al., 2004; Memeti et al., 2010; Frazer et al., 2014).

Inferred protracted construction of a vertically extensive San Juan batholith, where peak magma supply was intermittently focused at sites of ignimbrite eruptions and associated caldera collapse, leading to sustained maintenance of mushy granodioritic magma bodies in near-solidus environments, permits an interpretive crustal model of pluton assembly that is consistent with the volcanic eruptive history, regional geophysical data, geochronologic results, and petrologic evidence (Fig. 14). Early waxing magma supply fed eruptions at widely scattered sites in the upper crust, generating large central volcanoes dominated by intermediate compositions and only limited pluton volume in the upper crust. After several million years of dispersed magmatism, increasing magma supply, and warming of the upper crust, larger intrusive bodies could accumulate at shallow levels, focused at sites of closely clustered older volcanoes. Resulting caldera-scale granodioritic magma bodies could generate crystal-poor caps by low-pressure fractionation in the upper crust within a few hundred thousand years or less, leading to eruption of large ignimbrites. Continuation of such processes produced the overall composite batholith, 20–30 km thick, becoming more mafic downward. Geophysical definition of the ignimbrite-related batholith, but not for the earlier central volcanoes of the SRMVF, is closely analogous to the APVC in relation to the active volcanic zone farther west in the Andes.

Eruptible magma capable of sourcing ignimbrite eruptions accumulated only in the uppermost few kilometers (Fig. 14A), at local sites of focused high-mantle-magma supply. Other thick central parts of the San Juan batholith that lack associated calderas, such as the deep gravity low between the central and western caldera clusters (Fig. 2), probably record sites of sustained pluton assembly at lower rates without developing large volumes of eruptible magma. Deeper intrusive rocks, mainly granodiorite and tonalite similar to those exposed at midcrustal levels elsewhere in the Cordillera (southern Sierra Nevada, Coastal batholith of British Columbia), are likely at approximate density equivalence to wall rocks, thereby generating no detectible geophysical signature.

The seismically defined upper-crustal zone of high-velocity material (8–10 km depth; Fig. 3B), inferred from reprocessing of vintage seismic data (Drenth et al., 2012), is of unclear significance. It may represent pods of mafic batholithic rock, perhaps shallow cumulates that are complementary to erupted rhyolitic ignimbrites (Deering et al., 2011; Bachmann et al., 2014; Gelman et al., 2014). Alternatively, it might record a zone of mafic injection, similar to that inferred from seismic and deformation data, at slightly greater depth near Socorro, New Mexico (Sanford, 1983; Fialko and Simons, 2001), or seen deep in upper-crustal Cordilleran plutons (e.g., Best, 1963; Coleman et al., 1995; Miller et al., 2011). Because seismic modeling would be unable to resolve steep contacts between plutons and wall rocks, other interpretive complexities may exist.

Granitic rocks of the San Juan batholith are inferred to overlie an approximately equal thickness and volume of dense residua, including restite from widespread partial melting of lower crust along with cumulates from crystallization of voluminous mantle-derived mafic magma (assimilation-fractional crystallization [AFC] and melting-assimilation-storage-homogenization [MASH] processes; DePaolo, 1981; Hildreth and Moorbath, 1988). The modeled residua (Farmer et al., 2008) are too voluminous to be accommodated above the seismic Moho, within the present-day crust (which is relatively thin and low density; Prodehl and Lipman, 1989; Hansen et al., 2013). Either lower portions have delaminated or otherwise separated from the crust (as widely proposed for roots of Cordilleran-type magmatic systems; e.g., Arndt and Goldstein, 1989; Kay and Kay, 1993; Saleeby et al., 2003; Zandt et al., 2004; Jones et al., 2004; Jagoutz and Schmidt, 2012), or perhaps they have just become so dense that no geophysical distinction is possible between the residua and adjacent older mantle. Whatever the fate of these deep residua, the mid-Tertiary magmatic processes must have caused major chemical and physical reconstruction of the lithospheric column, probably accompanied by asthenospheric input as well (e.g., Farmer et al., 2008).

Assembly of Cordilleran Plutons

Processes of incremental pluton assembly, by recurrent addition of subhorizontal magma lenses, are well documented for some laccoliths and isolated plutons of modest size that are exposed at optimum crustal levels (Wiebe et al., 2002; Miller and Miller, 2002; Michel et al., 2008; Horsman et al., 2009; Rocchi et al., 2010; Miller et al., 2011). Such structural and compositional evidence of incremental assembly is unlikely to survive, however, in longer-lived and larger intrusive systems constructed by recurrent open-system recharge at high magma-supply rates, such as inferred for the San Juan batholith. Extensive upward (and downward?) flow late during pluton emplacement (or by convection in the reservoir; Gutierrez et al., 2013), as recorded by widespread steep mineral foliations in eroded Cordilleran plutons, would have disrupted and largely obliterated the early history of tabular magma assembly. Large individual plutons and composite batholith-size bodies, even if incrementally constructed by magma batches having lenticular aspect ratios, aggregate on scales that eventually occupy much of the crust. Nevertheless, although vertically extensive, typical composite Cordilleran batholiths, such as that beneath the San Juan region, retain an overall tabular “megalaccolithic” geometry due to great lateral extent (Fig. 14B).

Direct evidence for large volumes and prolonged life spans of batholith-scale mushy magma beneath loci of ignimbrite volcanism comes from the APVC in the central Andes (Ward et al., 2014). Anomalously slow seismic-shear velocities indicate that partial melt is currently present at 10–30 km depths beneath the entire surface footprint of the APVC (∼200 km diameter). The slow velocities (<2.9 km/s, locally to 1.9 km/s) would not be possible from even extreme variations in rock composition or temperature; a low-end estimate is 5% partial melt, locally possibly as much as 25%. The total volume of the mush zone beneath the APVC, estimated at 300,000 km3, cannot represent a single magma body; it must be a composite of multiple smaller zones of partial melt that amalgamated incrementally. Although individual calderas have no resolvable velocity expression, substantial lateral and vertical variations in melt percentage are likely as functions of volcano age, magma composition, and presence of wall-rock septa between plutons.

Maintenance or reinvigoration of such a large mush zone must depend on voluminous sustained recharge. Most plausibly, this involves processes at multiple crustal levels. Mafic mantle melts interact with the lower crust by AFC/MASH processes to generate intermediate-composition magmas with compositions much like the andesite and dacite that commonly erupt during early stages of an ignimbrite cycle. Despite clear isotopic and other petrologic evidence for involvement of enormous volumes of mantle-generated basaltic magma in the SRMVF and other Cordilleran systems, little if any basaltic magma reached the surface centrally within areas of high mid-Tertiary magmatic fluxes marked by ignimbrite calderas and geophysically constrained subvolcanic intrusions.

Rising basaltic and derivative intermediate-composition melts would mix efficiently with residual mush of an upper-crustal magma body that differs only modestly in composition and density, thereby increasing temperatures and melt proportion in the mush zone. Such processes would permit assembly of an eruptible melt-rich zone at the top of the mush body (as commonly diagrammed for individual calderas: e.g., fig. 7 inHildreth, 2004; fig. 6 inde Silva et al., 2006), along with smaller-volume melt lenses at greater depth (Cashman and Giordano, 2014). The continued presence of voluminous mushy magma beneath the long-lived APVC (10 m.y.) implies prolonged cycles of pluton assembly and crystallization, which should have counterparts in the durations (to 106 m.y.) of zircon ages within single plutons, and even within individual samples from the Sierra Nevada and other batholiths. Two-dimensional thermal models for Sierran plutons are also consistent with assembly over intervals as long as several million years (Paterson et al., 2011).

The presence of laterally extensive bodies of interconnected melt high in such long-lived mushy plutons is recorded by the eruption of large calc-alkaline ignimbrites (102–103 km3; especially crystal-poor rhyolites and those that grade into late-erupted dacite with more complex crystal cargos; Hildreth, 1981; Bachmann and Bergantz, 2004), triggering multikilometer caldera collapse bounded by ring faults. Field data document common subsidence depths of 2–5 km at such calderas (Lipman, 1984; John et al., 2008; Best et al., 2013), confirming large integrated volumes of eruptible magma at shallow depth; ignimbrite and subsidence volumes are broadly proportional (Smith, 1979; Spera and Crisp, 1981; Gregg et al., 2012; Geshi et al., 2014). Analogue models show that the geometry of caldera subsidence changes from ring-fault to downsag depression as thickness of the roof above the magma increases (Marti et al., 1994; Roche et al., 2000; Acocella et al., 2000; Kennedy et al., 2004), but large downsag calderas that should result from draining of deep magma rarely accompany voluminous ignimbrite eruptions (Lipman, 1997; Cole et al., 2005). Other silicic magma bodies that are sufficiently liquid to erupt, especially as sources for smaller ignimbrites with disequilibrium crystal assemblages, may develop as multiple poorly connected melt pockets at more varied depths, which become interconnected and mixed shortly before or during the eruption (e.g., review by Cashman and Giordano, 2014).

Some large ignimbrites record much briefer zircon-crystallization histories than the prolonged magmatic evolution inferred here from the U/Pb age spectra in Cordilleran arcs. For example, recent TIMS analyses from several relatively dry, high-temperature rhyolite ignimbrites of the Yellowstone–Snake River Plain region (Huckleberry Ridge, Kilgore Tuffs) indicate that most zircons in these intraplate magmas crystallized within 5000–10,000 yr prior to their eruption, as limited by 40Ar/39/Ar ages (Rivera et al., 2014; Wotzlaw et al., 2014; Bindeman and Simakin, 2014). The tight grouping of zircon ages for these ignimbrites, without xenocrysts recording Archean basement and near absence of antecrysts from earlier volcanic activity, strongly suggests that these magma batches remained mostly zircon-undersaturated and able to resorb assimilated zircon crystals until late in their assembly. Any antecrystic zircons inherited from early stages in magma generation and assembly, involving voluminous assimilation of hydrothermally altered country rocks as inferred from low 18O values of the Kilgore magmas (Bindeman et al., 2007; Wotzlaw et al., 2014), are poorly recorded by the dominant zircon ages. In contrast, low 18O values are largely or entirely absent in the colder-wetter, calc-alkaline ignimbrite magmas of the SRMVF and other mid-Tertiary ignimbrites of the Cordilleran arc (Larson and Taylor, 1986), indicating limited involvement of altered upper crust in their generation. Because early-formed antecrystic zircons in continental-arc ignimbrites would be less soluble than in the hot-dry silicic magmas of the Snake River Plain (e.g., Watson and Harrison, 1983; Miller et al., 2003), they preserve lengthier records of magma-body assembly: e.g., 500–1000 k.y. for the Fish Canyon Tuff (Wotzlaw et al., 2013; Coble, 2013), and 250–500 k.y. for the younger Taupo, Toba, and Kos Plateau ignimbrites (Brown and Fletcher, 1999; Charlier et al., 2005; Reid, 2008; Guillong et al., 2014).

Geologic, geophysical, and geochronological data are consistent with prior proposals that large, long-lived, silicic volcanic fields such as the SRMVF are surface expressions of composite upper-crustal magma bodies comparable to the Boulder or Sierra Nevada batholiths. Recent data summarized here permit improved quantitative assessment of batholith geometry, duration of magma assembly and crystallization, and rates of magma supply during evolution of the SRMVF and comparable Cordilleran magmatic systems.

The San Juan volcanic locus, associated batholith, and the broader SRMVF record multimillion-year incremental construction by prolonged open-system processes at high average rates of magma supply, involving voluminous mafic-mantle inputs, large-scale crustal assimilation, and concurrent generation of dense residua (cumulate, restite) that now lie mostly beneath the seismic Moho. Recurrent generation of melt-dominated silicic magma lenses at times of peak magma supply, above vertically extensive bodies of near-solidus crystal mush in the upper crust and underlain by voluminous input from mafic mantle-derived magma, led to repeated ignimbrite supereruptions and caldera formation in the SRMVF. Erupted SRMVF rhyolites have liquid compositions indicative of low-pressure fractionation; most dacites carry disequilibrium crystal-rich cargos but have a low-pressure rhyolitic groundmass. Complementary crystal accumulations resulting from rhyolite fractionation have been identified in some systems, although they are commonly obscured by high volumetric ratios of vertically extensive intrusions relative to extracted evolved liquid and the intrinsic inefficiency of extracting highly viscous melt from crystalline residue. The compositions of some voluminous eruptions in the SRMVF varied widely within time spans too brief to resolve by available geochronologic methods (e.g., the San Luis caldera complex; Lipman and McIntosh, 2008). In contrast to the protracted evolution of the overall volcano-plutonic system, eruptible ignimbrite magmas with high melt proportions must have been generated rapidly, with only brief lifetimes before venting. The bulk of eruptible magma reached the surface as volcanic deposits, but the more voluminous underlying magma mush continued to evolve and to be recharged by sustained mantle input, finally solidifying only when deep magma supply waned. The volume of the overall mantle input likely was 1–2 times that of the volcanic ejecta and underlying crustal batholiths (Fig. 4).

Recently published CA-TIMS zircon ages for SRMVF intrusions (Figs. 7–8) define average times of crystallization as the magmatic system cooled but do not document the total duration of pluton assembly. Mineral ages (mainly biotite, hornblende) determined by 40Ar39Ar methods are only slightly younger than the zircon ages, indicating that late crystallization was accompanied by rapid cooling and solidification. Crystallization ages for caldera-related intrusions vary from indistinguishable from the associated ignimbrite to ∼2 m.y. younger, and age differences tend to increase with depth of pluton emplacement (Fig. 10). Inferences that such age variations date times of intermittent magma assembly are not supported by any documented age-depth correlation within a single pluton. More plausibly, deeper levels of plutons that cooled slowly preserve a lengthier record of ascent and crystal growth in sequential magma pulses, punctuated by periods of zircon undersaturation and crystal resorption (Fig. 10B). Age variations among separate crystals from a single sample and among multiple samples from deeper intrusions in the Sierra Nevada are much larger than for the subvolcanic plutons in the Southern Rocky Mountains (Fig. 8), a relation consistent with more protracted pluton construction and crystallization at greater depth.

Proximal intrusions of focused magmatism in the SRMVF were vertically extensive and capable of supporting large ignimbrite eruptions, as high magma supply mixed and variably homogenized successively emplaced incremental inputs. In contrast, more-distal plutons tend to be smaller and laccolithic in shape, and they may more closely record assembly from successive magma increments that largely crystallized before the next intrusive pulse. Near-absence of Proterozoic xenocrysts in SRMVF caldera-related intrusions, in contrast to their abundance in some peripheral intrusions (Fig. 9), suggests lengthier assembly histories for the caldera intrusions, including intervals of zircon undersaturation and dissolution. Varied crystallization ages laterally and vertically across a pluton are inferred to reflect existence of discontinuous domains where magma composition and pressure-temperature conditions were locally conducive to zircon saturation and crystal growth at different times. Large intra- and intergrain variations in concentrations of U, Hf, and other trace elements (to order-of-magnitude) from grain to grain and zone to zone in zircons and within a single intrusion sample demonstrate variable growth environments at many scales within small magma batches (Schaltegger et al., 2009; Claiborne et al., 2010; Zimmerer and McIntosh, 2012a, their supplemental data; Erdmann et al., 2013). Ephemeral local compositional gradients must have developed during crystallization and late-stage mixing.

Subtle compositional, textural, and structural discontinuities within lithologically mappable pluton phases also suggest the existence of spatially discontinuous domains during prolonged assembly and sustained maintenance of crystal mush through much of an evolving pluton. Individual or small groups of rising magma pulses, which cool to form discrete plutons in the shallow crust, merge and coalesce with adjacent residual crystal mush at deeper levels to form larger granitoid bodies in which boundaries between magma packages become gradational or otherwise obscure. Extraction of interstitial melt, heterogeneous flowage, and mixing within remaining crystal mush could further complicate the distribution of apparent ages within and between such domains in a pluton.

Accordingly, contrary to some interpretations, zircon U/Pb ages provide few constraints on processes or durations of magma-body assembly in granitoid plutons, especially at times of high magma supply. Prolonged duration of pluton construction and mixing of crystal cargo late during assembly are documented by SIMS age ranges of >105 yr within zoned single zircon crystals, similar spans among CA-TIMS single-crystal zircon ages from individual samples of caldera-related intrusion, and antecryst zircon ages that predate eruption age of the associated ignimbrite. Close agreement among autocryst zircon and titanite ages, and only slightly younger 40Ar/39Ar hornblende dates suggest major zircon growth late during assembly of granitic plutons as the magma supply wanes and cooling accelerates. During earlier stages, compositionally contrasting intrusive phases within a composite pluton complex can have coexisted for lengthy periods as largely crystallized bodies with residual interstitial liquid, subject to successive additional magma pulses that could slow or reverse cooling and affect zircon stability.

Based on vertically extensive intrusion geometry, total magma supply for many Cordilleran batholiths (>0.005 km3/yr for the overall intrusion assembly at the San Juan locus) would have been sufficient to maintain large volumes of crystal mush above the solidus for multimillion-year durations in the upper crust (Fig. 12). Pluton-fill rates in such composite batholiths likely fluctuated substantially both in time and space; times of high supply (likely >0.01–0.05 km3/yr) focused at localized areas would generate large-volume (102–103 km3) capping lenses of eruptible magma to form ignimbrite eruptions and associated calderas. The inferred vertically extensive magma-body geometry, interaction of voluminous mafic magma from the mantle with crustal melts, prolonged durations of pluton assembly and crystallization, and voluminous silicic volcanism occurring concurrently with batholith construction as recorded by the SRMVF are suggested to typify continental-margin arc magmatism globally.

A stimulating Geological Society of America (GSA) Field Forum on the Sierra Nevada batholith, organized in 2005 by Drew Coleman, Allen Glazner, and John Bartley, re-energized Lipman’s interest in issues of pluton-crystallization ages and duration of magma-body assembly. Subsequent studies on subvolcanic plutons in the SRMVF (by Coleman and students at the University of North Carolina, Matt Zimmerer and Bill McIntosh at New Mexico Tech, and David Gonzales at Fort Lewis College) generated enjoyable field work, thought-provoking data, and some divergent interpretations. We especially thank David Gonzales and Kathryn Watts for sharing data on zircon populations in western San Juan intrusions and the Caetano caldera system. Jake Lowenstern and Kathyrn Watts of the U.S. Geological Survey, and Geosphere reviewers and editors (Ben Drenth, Jonathan Miller, Lang Farmer, and Shan de Silva) provided exceptionally helpful comments on the manuscript. This work was partly supported by Schweizerischer Nationalfonds (SNF) fund 200021_146268 to Bachmann.