An ignimbrite caldera from the bottom up: Exhumed floor and fill of the resurgent Bonanza caldera, Southern Rocky Mountain volcanic field, Colorado

The composite Southern Rocky Mountain volcanic field (SRMVF) (Fig. 1) has long been studied as a site of mid-Tertiary silicic volcanism on especially voluminous scales (Cross and Larsen, 1935; Larsen and Cross, 1956; Lipman et al., 1970; Epis and Chapin, 1974; Steven and Lipman, 1976; McIntosh and Chapin, 2004), including at least 28 ignimbrite sheets (each 150–5000 km³) and associated calderas active at 37–23 Ma (Tables 1 and 2). Ignimbrite-caldera systems of the San Juan Mountains, constituting the largest preserved erosional remnant of the SRMVF, have been a special focus for recent volcanologic and petrologic research: southeastern calderas (Platoro complex: Dungan et al., 1989; Lipman et al., 1996), western (Uncompahgre-Silverton–Lake City: Hon and Lipman, 1989), and the central cluster (La Garita–Creede calderas: Lipman, 2000, 2006; Bachmann et al., 2002, 2007). Much less examined have been Bonanza and adjacent Marshall calderas in the northeast San Juan region (Figs. 2 and 3), which define a transition from earlier volcanism in central Colorado to the larger-volume younger ignimbrite-caldera foci farther southwest (Fig. 1). Other than mineral-resource studies of small areas (e.g., Scott et al., 1975; Van Alstine, 1975; Olson, 1988), most of the northeastern San Juan Mountains until recently had been examined only in reconnaissance for the Colorado State geologic map (Tweto et al., 1976; Tweto, 1979).

Existence of a caldera has long been inferred in the Bonanza area, based on gravity data (Karig, 1965) and regional reconnaissance studies (Steven and Lipman, 1976; Varga and Smith, 1984), but detailed geologic mapping, petrologic information, and geochronologic data have been sparse. The only previously published geologic maps for any part of Bonanza caldera were the pioneering report on the Bonanza mining district by Patton (1916) and a more detailed study of the district at a scale of 1:12,000 (Burbank, 1932). These studies distinguished the major local rock units for a relatively small area and provided information on mine workings, but both were undertaken before development of modern concepts for ignimbrite volcanism and associated caldera subsidence. Several theses at the Colorado School of Mines in the late 1960s and early 1970s, which focused on the Bonanza area as a “ground-truth” test area for remote-sensing data, provided lithologic information of varying detail for areas adjacent to the mining district (Bridwell, 1968; Kouther, 1969; Mayhew, 1969; Knepper and Marrs, 1971; Perry, 1971; Marrs, 1973). These studies included recognition that the unit named Bonanza latite and interpreted as a thick lava sequence by Patton (1916), who carefully described the presence of abundant small fragments of andesite and puzzled over its “rhyoclastic” texture, consisted of welded tuff (Bruns et al., 1971).

Concurrently, a detailed stratigraphy for outflow ignimbrites (ash-flow tuffs) was developed in the Thirtynine Mile volcanic area farther to the northeast (Fig. 1; Chapin and Epis, 1964; Epis and Chapin, 1968, 1974), but with insufficient regional control to locate eruptive sources unambiguously. More recent geochronologic and paleomagnetic studies of the distal ignimbrites in the Thirtynine Mile area, and comparisons with intracaldera ignimbrites and associated rocks along the Sawatch trend as far south as Bonanza, demonstrated that one of the eastern ignimbrites, the Gribbles Park Tuff of Epis and Chapin (1974), is indistinguishable in age and paleomagnetic direction from proximal Bonanza Tuff (McIntosh and Chapin, 2004). Despite these discoveries, uncertainties have continued about stratigraphic and structural relationships in the Bonanza area and about the mid-Tertiary ignimbrite sources (Grizzly Peak, Princeton, and Aetna) that are aligned along the Sawatch Range to the north (Fig. 1; Shannon, 1988; Fridrich et al., 1991; McIntosh and Chapin, 2004).

Our study of the Bonanza area evaluates eruptive and magmatic processes of silicic Cordilleran volcanism, based on new geologic mapping (mainly summers of 2007–2011), high-precision ⁴⁰Ar/³⁹Ar age determinations (130 localities; 138 mineral and groundmass ages: Fig. 4, Table 3, Supplemental Tables 1¹, 2², and 3³), and chemical and petrographic data including new major-oxide and trace-element analyses for ∼280 samples (Table 4, Supplemental Table 4⁴). Methods are summarized in the Appendix and Supplemental Tables. Although surface exposures are not exceptional (heavy vegetation and widespread talus cover on some steep slopes), the Bonanza center displays unusually complete and diverse features of a large ignimbrite caldera cycle. These include voluminous andesite and more silicic lavas erupted before the ignimbrite eruptions, complex compositional zonations within both the outflow sheet and tuff concurrently ponded within the caldera as a single ignimbrite unit, extensive portions of the ring-fault system that accommodated caldera subsidence, thick compositionally diverse lavas that filled the caldera after subsidence, eroded remnants of the original topographic caldera rim, widespread exhumation of caldera-floor features, and postcaldera intrusions that core a notably steep resurgent dome within the caldera (Fig. 5).

Among broader topics examined are some complexities of pyroclastic eruption and emplacement, geometric relations between caldera subsidence and resurgence, petrologic diversity of sequential eruptions that formed a single ignimbrite, volumes of outflow and intracaldera ignimbrite in relation to caldera size and inception of subsidence, recurrent eruption of intermediate-composition lavas after caldera-forming events, emplacement of subvolcanic plutons, magnitude and rates of caldera resurgence, relations to regional extensional faulting, involvement of mantle-derived mafic components in magma generation, time-space-volume-compositional progressions in the SRMVF, and comparisons with continental-margin volcanism elsewhere.

Mid-Tertiary volcanic deposits once were continuous across much of the Colorado and northern New Mexico mountains (Fig. 1), constituting the composite Southern Rocky Mountain volcanic field (SRMVF), for which the San Juan region is the largest erosional remnant (Steven, 1975; Lipman, 2007). Subareas of the SRMVF, dismembered by subsequent erosion, have commonly been described, somewhat misleadingly, as separate volcanic fields (San Juan, Sawatch, Thirtynine Mile, Latir, West Elk, and Central Colorado), rather than as time-space transgressive magmatic foci within a large composite field.

Activity in the SRMVF peaked between 37 and 26 Ma (McIntosh and Chapin, 2004; Lipman, 2007; Lipman and McIntosh, 2008). Dominantly intermediate-composition lavas and breccias (andesite-dacite), erupted from widely scattered central volcanoes, were characteristic early phases of SRMVF activity. Major volcanic foci, initially established by clustered stratocones, became eruptive sites for ∼28 caldera-associated ignimbrites of more silicic compositions (Table 1 and Fig. 6), in response to increased magmatic input. Composite volumes of the early-intermediate volcanoes are large; in the San Juan region, stratigraphic sequences of the early volcanic rocks commonly are more than a kilometer thick, and total volume (∼25,000 km³; Lipman et al., 1970) is nearly twice that of the later-erupted ignimbrites.

The original areal extent of the overall SRMVF appears to have exceeded 100,000 km², with a total volume of volcanic deposits greater than 60,000 km³ (Lipman, 2007; Lipman and Bachmann, 2015). Peak magmatic volumes in the SRMVF, associated with ignimbrite eruptions, define a general (if imperfect) progression, from early eruptions along the trend of the Sawatch Range in central Colorado (37–34 Ma), southward into the San Juan region (33–27 Ma), and later to the 25-Ma Latir-Questa locus in northern New Mexico and the 23-Ma Lake City caldera in the western San Juan Mountains (Fig. 6). Geophysical data document the presence of several composite subvolcanic batholiths that encompass most calderas of the SRMVF (Plouff and Pakiser, 1972; Drenth et al., 2012; Lipman and Bachmann, 2015). The SRMVF is among several discontinuous loci of intense Tertiary volcanic activity—including the Sierra Madre Occidental, Trans-Pecos, Mogollon-Datil, Absaroka, Challis, and Lowland Creek fields—that developed along the eastern Cordilleran margin of the North American plate, probably in a complex response to changing subduction dynamics along the western plate boundary (Lipman, et al., 1972).

Eruptions beginning in northern parts of the SRMVF at least as early as 38 Ma (Epis and Chapin, 1974; McIntosh and Chapin, 2004) are the precursor framework for ignimbrite volcanism in the Bonanza area. Caldera sources for the early (37–33 Ma) large ignimbrites (Table 1) are aligned north-northwesterly along the crest of the Sawatch Range (Fig. 1), but preserved exposures of the outflow tuff sheets are limited, especially west of the present-day Continental Divide. The stratigraphy, regional distribution, and eruptive history of these ignimbrites remain incompletely documented; main features are summarized briefly here.

The first major regional ignimbrite was the far-traveled Wall Mountain Tuff (Chapin and Lowell, 1979), a relatively crystal-rich rhyolite (Table 2). It erupted at 36.9 Ma, probably from a now completely eroded caldera above the 25 × 35-km Mount Princeton batholith (Fig. 1 and Table 1: Shannon, 1988; Lipman and Bachmann, 2015; alternative interpretation in Mills and Coleman, 2013). Distal Wall Mountain Tuff is preserved on the High Plains, more than 150 km east from its source (Chapin and Lowell, 1979). In contrast, a few small erosional remnants northwest of Cochetopa Park (Fig. 1), 70–80 km southwest of Mount Princeton, are the only known localities west of the Sawatch Range for this large ignimbrite (Lipman and McIntosh, 2008; Lipman, 2012).

North of the Mount Princeton batholith, the spectacularly exposed Grizzly Peak caldera and associated intracaldera tuff (Fridrich et al., 1991) formed at 34.3 Ma (McIntosh and Chapin, 2004), but the presumed equivalent outflow tuff sheet has been almost completely eroded. Nested within the southern margin of the Princeton batholith is the slightly younger Mount Aetna caldera (Shannon, 1988), source of the 34.1-Ma Badger Creek Tuff (Epis and Chapin, 1974; Shannon et al., 1987) and associated intrusions (34–29.6 Ma; Zimmerer and McIntosh, 2012; Mills and Coleman, 2013). At the south end of the Sawatch trend, Marshall caldera (within the Bonanza area: Fig. 2) was the source for the 33.9-Ma Thorn Ranch Tuff that is preserved mainly in the Thirtynine Mile region east of the Rio Grande rift (Epis and Chapin, 1974; McIntosh and Chapin, 2004).

During the present study, it became clear that two previously described and named regional tuffs in the Thirtynine Mile area, the East Gulch and Stirrup Ranch Tuffs of Epis and Chapin (1974; McIntosh and Chapin, 2004), are not large discrete ignimbrites, thereby increasing the recurrence intervals between early SRMVF eruptions (Fig. 6). The previously described East Gulch Tuff (Epis and Chapin, 1974) is the non-welded to weakly welded basal cooling zone of the Thorn Ranch Tuff, as clearly exposed at several sections, including sites sampled for geochronologic and paleomagnetic determinations (McIntosh and Chapin, 2004). The transition upward into the densely welded cliff-forming interior of the Thorn Ranch Tuff is completely gradational, and the isotopic ages and paleomagnetic pole directions are indistinguishable. The Stirrup Ranch unit is a local sheet of coarse monolithologic breccia, consisting entirely of angular clast-supported blocks, with only minor interstitial matrix of comminuted tuff. All the blocks are densely welded Wall Mountain Tuff, as identified by similar rock chemistry, phenocryst populations, and ⁴⁰Ar/³⁹Ar ages. The absence of bedding and random orientation of large clasts (to several meters) in this unit are closely comparable to some monolithologic welded-tuff breccias in central Nevada that have been interpreted as products of catastrophic dam-burst–type floods resulting from short-term blockage of paleovalleys by volcanic rocks or landslides (Henry, 2008), and a similar origin seems likely for the Stirrup Ranch deposit.

The San Juan Mountains are the largest erosional remnant of the SRMVF (Fig. 1). Preserved volcanic rocks occupy an area of more than 25,000 km² and have a volume of ∼40,000 km³. They cover a varied basement of Precambrian to early Tertiary rocks along the uplifted and eroded west margin of the Late Cretaceous to early Tertiary (Laramide) uplifts of the Southern Rocky Mountains and adjoining eastern parts of Colorado Plateau (Fig. 1). As mid-Tertiary volcanism migrated southward from the Sawatch Range (Fig. 6), widely scattered intermediate-composition centers erupted lavas and flanking volcaniclastic breccias starting at 35–34 Ma (Lipman et al., 1970; Lipman and McIntosh, 2008). These rocks, which constitute about two-thirds the volume of the preserved volcanic assemblage, are widely overlain by large ignimbrites associated with caldera collapses (Steven and Lipman, 1976).

After ignimbrite eruptions from Marshall and Bonanza calderas at 33.9 and 33.1 Ma in the northeastern San Juan Mountains (Fig. 6 and Table 3), magmatic activity migrated southwest with eruption of the Saguache Creek Tuff from the North Pass caldera at 32.2 Ma (Lipman and McIntosh, 2008; Lipman, 2012), then to the southeast San Juan region at 30.1 Ma (inception of Platoro caldera complex), followed shortly by eruptions mainly of crystal-poor rhyolitic ignimbrites from western calderas (Steven and Lipman, 1976). Ignimbrite activity progressively focused in the central San Juan Mountains (Tables 1 and 2), leading to eruption of the enormous Fish Canyon Tuff (5000 km³ of monotonously uniform crystal-rich dacite) and collapse of the 35 × 75 km La Garita caldera at 28.0 Ma (Lipman, 2000, 2006; Bachmann et al., 2002). In the central San Juans, seven more eruptions of compositionally diverse ignimbrite, with volumes of 100–1000 km³, erupted during the 1.1-m.y. interval from 28.0 to 26.9 Ma from calderas nested within La Garita caldera (Figs. 1 and 2). At ca. 26 Ma, magmatism shifted to a bimodal assemblage dominated by trachybasalt and silicic rhyolite, concurrent with the inception of regional extension along the Rio Grande rift.

The Bonanza area is more complex structurally than most other sectors of the San Juan Mountains: (1) deep erosion has exposed prevolcanic structures and paleotopography in Paleozoic and Proterozoic rocks beneath the thick Tertiary volcanic cover of the SRMVF; (2) ring-fault subsidence and uplift structures of the strongly resurgent Bonanza caldera are exposed at deep levels down to the floor of this geometrically complex caldera; (3) widespread alteration and vein mineralization obscure stratigraphy and structure within the Bonanza mining district; and (4) postvolcanic extension has produced normal faults, stratal tilting, and accommodation-zone structures at the junction between the San Luis Valley and Upper Arkansas rift segments of the Rio Grande rift zone (Fig. 3). Recognition and interpretation of the multiple episodes of faulting and other structures are further hindered by heavy vegetation and limited outcrops in many areas, by lack of well-defined stratigraphy within the massive ignimbrites filling the Bonanza and Marshall calderas, by uncertain distinctions among lava sequences of multiple ages that have similar lithologies, and by difficulties in distinguishing effects of fault offsets from stratigraphic discontinuities resulting from non-planar deposition of volcanic units in deep paleovalleys. Several unconformities along flanks of large paleovalleys in the Bonanza area previously have been represented on regional maps as large faults (e.g., Tweto, 1979). In the present study, faults have been depicted only where evidence for displacement is clear; some unmapped structures undoubtedly exist in places where field relations are inadequately convincing.

Mid-Tertiary volcanic rocks of the Bonanza area lie along the broad boundary between Precambrian-cored uplifts of the Southern Rocky Mountains and less deformed Paleozoic and Mesozoic sedimentary rocks along the northeast margin of the Colorado Plateau. These contrasts in geologic setting exerted important controls on depositional, structural, and morphologic evolution. To the south and west in the San Juan region, gently dipping Mesozoic sedimentary rocks semi-conformably underlie the volcanic sequence, and the ignimbrite sheets form a stratified plateau, interrupted mainly by local structures associated with caldera and central-volcano constructs. The original mid-Tertiary volcanic terrain was much like the Altiplano of the central Andes (de Silva et al., 2006), because voluminous eruptions buried and subdued much of the preexisting topography.

In contrast, volcanic accumulations in the Bonanza area lap onto rugged paleotopography associated with the Late Cretaceous to early Tertiary (Laramide) uplifts. Large paleovalleys, which developed during erosion of the early Tertiary uplifts and during the growth of central volcanoes prior to ignimbrite eruptions, strongly influenced the distribution of subsequently emplaced volcanic deposits (e.g., Chapin and Lowell, 1979; Steven et al., 1995). Some major paleovalleys survived the entire period of volcanism, and many present-day drainages are inherited from the mid-Tertiary landscape, leading to our fieldwork expression “once a valley, always a valley.”

Pre-Tertiary rocks also provide useful information on paleotopography at the inception of volcanism and subsequent events in the Bonanza area. Notably, Precambrian basement and small windows of Paleozoic sedimentary rocks, but no Mesozoic rocks, are exposed around margins of the Marshall and Bonanza calderas, showing that they formed within a prevolcanic highland, probably a southern continuation of a Late Cretaceous uplift in the vicinity of the present-day Sawatch Range. Farther west, gently dipping Mesozoic sedimentary rocks are widely present beneath the volcanic cover (Tweto, 1979; Lipman, 2012), reflecting proximity to the Colorado Plateau. Absence of early Tertiary sedimentary deposits along the unconformity at the base of the volcanic sequence in the map area, such as preserved farther south and west in the San Juan region (Blanco Basin Formation and Telluride Conglomerate), is further evidence that the prevolcanic paleosurface stood topographically high and was primarily a region of erosion rather than deposition.

Due to the greater paleorelief, the stratigraphic record of sequential eruptions is less complete in the Bonanza area than to the southwest. Many volcanic deposits in the northeast sector accumulated in broad valleys, which were incompletely filled and then re-excavated by erosion between successive eruptions. The regional ignimbrites, rather than forming a stratified plateau, in places are preserved in inverted topographic order, with earlier tuff sheets capping ridges and younger units exposed at lower levels within paleovalleys. In many places, welded tuffs are preserved as isolated scabs, unconformable against slopes of paleovalleys, without stratigraphic continuity between sequential deposits. Frequent miscorrelations of ignimbrite units in previously mapped parts of the northeastern San Juan region resulted from such complexities, as well as from limited exposures due to forest cover, incomplete knowledge of the regional eruptive sequence, and inadequate recognition of petrographic distinctions among tuff sheets.

Proterozoic and Paleozoic rocks that crop out widely around eastern and northern margins of the Bonanza caldera define a rugged paleosurface of highlands and valleys that were repeatedly partially buried and re-excavated during mid-Tertiary volcanism. The Proterozoic rocks include large areas of coarsely porphyritic granodiorite, intruded into diverse metasedimentary and metavolcanic country rocks of quartzo-feldspathic gneiss, schist, and amphibolite.

Paleozoic sedimentary rocks preserved between Precambrian basement and overlying Tertiary volcanic rocks provide important evidence for pre- and synvolcanic structural events. Lower Paleozoic strata are dominantly marine carbonate deposits (Burbank, 1932; Cappa and Wallace, 2007); these have long been subdivided into several formations that provide high-resolution structural markers. Upper Paleozoic rocks are thick continental clastic deposits that record concurrent uplift (Ancestral Rocky Mountains) and erosion; distinctive subunits are lacking, and detailed subdivision has not been possible.

The pre-Tertiary rocks were further deformed, uplifted, and deeply eroded during Laramide compressional growth of the Rocky Mountains in the Late Cretaceous and early Tertiary, and the mid-Tertiary volcanic rocks were deposited mainly on Precambrian rocks with only local surviving remnants of the Paleozoic sequence. The largest areas of exposed Paleozoic strata, on both sides of lower Kerber Creek, define a south-plunging anticline and an E-W–trending zone of faulting (Fig. 3). These features have been interpreted as Laramide structures (Burbank, 1932, p. 37–42; Cappa and Wallace, 2007, p. 38, 45), but (below) we infer that they largely resulted from subsidence and subsequent resurgence of Bonanza caldera.

As elsewhere in the San Juan region, ignimbrite sheets and other rocks associated with calderas in the Bonanza region overlie thick lava sequences erupted from large central volcanoes (Figs. 2 and 6), recording initial assembly of upper-crustal magma bodies preparatory to caldera-scale explosive eruptions. Sections of the early lavas as thick as 2.3 km have been penetrated by petroleum exploration drilling southeast of the Bonanza area (Gries, 1985; Brister and Gries, 1994). In comparison with the early-intermediate assemblage farther south and west in the San Juan Mountains, lava thicknesses in the northeastern area are more variable due to the rugged prevolcanic paleotopography. The assemblage in the Bonanza region also contains greater proportions of proximal lavas and breccias relative to distal laharic conglomerates and other volcaniclastic rocks, and dacite and rhyolite are more voluminous components of the dominantly andesitic lava assemblage.

Rocks of the central volcanoes are broadly correlative with the regional Conejos Formation (Lipman et al., 1970; Colucci et al., 1991), but several early-erupted tuff sheets from calderas along the Sawatch trend are interstratified with lavas that predate all ignimbrites from farther south and west in the San Juan Mountains (Fig. 7). The ignimbrites and sparse sanidine-bearing lavas help define the eruptive history of early lavas in more detail than possible elsewhere in the region. The lavas that are interstratified with the early tuff sheets, as well as andesitic and dacitic lavas that ponded within ignimbrite calderas throughout the San Juan region, are compositionally indistinguishable from earlier-erupted lavas of the Conejos assemblage and document eruptive and compositional continuity during growth of the SRMVF.

On a regional scale, total thickness and volume of the early lavas are far greater than for interstratified and overlying ignimbrite sheets. In the Bonanza area, lavas are locally exposed over a vertical range of more than 1000 m from along Saguache Creek to the Continental Divide, but thickness tends to decrease northward toward the Gunnison Valley and eastward toward the San Luis Valley segment of the Rio Grande rift zone. Paleohills of basement rocks project through the volcanic cover more commonly than farther to the south and west in the San Juan region. Within the Bonanza caldera area (Fig. 2), the early lavas thin against prevolcanic paleohills, and in some paleovalleys, Bonanza Tuff was deposited directly against Proterozoic granite.

Parts of at least five major eruptive centers for pre-Bonanza lavas, all equivalent to rocks of the Conejos Formation farther southwest, lie within the Bonanza caldera area, and others are present farther west (Lipman, 2012). These centers probably include the earliest eruptions in the San Juan region (Table 3; Supplemental Table 1). Eruptions shifted west and south with time. The Jacks Creek volcano (34.6–34.2 Ma) and andesitic and dacitic lavas (33.9–33.4 Ma) that filled Marshall caldera (eruptive source of the Thorn Ranch Tuff) were followed by the Rawley cluster of volcanoes (33.7–33.3 Ma) that formed a volcanic highland within which Bonanza caldera is centered. A thick andesite pile of poorly constrained age (ca. 33.2 Ma?) accumulated along Sargents Mesa along the Continental Divide west of Bonanza, and the large Tracy volcano south of the Saguache valley (ca. 33.7–31.6 Ma) erupted intermediate-composition lavas both before and after the Bonanza Tuff.

Eruption of the Thorn Ranch Tuff from the previously unmapped Marshall caldera (Figs. 2 and 3) was the most proximal antecedent to the ignimbrite from Bonanza 0.8 m.y. later, and features of this caldera provide the basic stratigraphic and morphologic framework for interpreting the younger volcanic activity. The thick sequence of intermediate-composition lavas that filled Marshall caldera merge with precursor lavas of the Bonanza cycle, and Bonanza caldera caved away the southeast margin of the earlier caldera. Along with the slightly younger Saguache Creek Tuff (32.2 Ma; Lipman and McIntosh, 2008), these three major ignimbrite eruptions and associated calderas of the northeast San Juan Mountains bracket a major geographic transition, from earlier SRMVF activity along the Sawatch Range trend, into the locus of peak eruptive activity in the San Juan region (Figs. 1 and 5).

Probable presence of an ignimbrite caldera in upper Marshall Creek was initially inferred by Thomas Steven (1986, written commun.), based on reinterpretation of published geologic mapping in the Marshall Pass mining district (Olson, 1983). Olson’s detailed map depicts a lithologic assemblage (welded tuff interleaved with volcanic breccia, overlain by tuffaceous lake-bed deposits) similar to that within ignimbrite calderas elsewhere in the San Juan Mountains (Steven and Lipman, 1976). Subsequently, during study of fossil flora in lake sediments of the Pitch-Pinnacle Formation, Gregory and McIntosh (1996) dated the welded tuff in upper Marshall Creek, noted associated megabreccia of intracaldera type, and suggested that this assemblage marked a caldera source for the 33.9-Ma Thorn Ranch Tuff preserved widely in the Thirtynine Mile area to the northeast (Epis and Chapin, 1974; McIntosh and Chapin, 2004).

Our present study augments these interpretations, demonstrating that eruptions from Marshall caldera were major precursors to inception of the Bonanza eruptive cycle. Although largely concealed beneath younger lavas and partly truncated by Bonanza caldera, Marshall caldera is the oldest ignimbrite source in the San Juan region. Along the previously unrecognized northeast margin of this caldera, intracaldera Thorn Ranch Tuff, >400 m thick with no base exposed, banks against Proterozoic wall rocks at a geographic feature known as “The Gate” (Figs. 8 and 9), east of Marshall Pass. This steep unconformity has been depicted on regional maps as a contact of andesite lava against the Proterozoic basement (Tweto et al., 1976; Cappa and Wallace, 2007), but densely welded, lineated tuff dips parallel to caldera-wall contacts at The Gate and contains large fragments of Proterozoic rock derived from the wall. Dips in the Thorn Ranch Tuff decrease, from locally near-vertical at the caldera wall, to as low as 20°–25° at 100–200 m distance from the contact.

The northern wall of Marshall caldera is exposed continuously from The Gate northwestward into upper Marshall Creek (Figs. 3 and 9), where it projects southward, as constrained by exposed Proterozoic and Paleozoic rocks farther west. Its southwestern margin is concealed beneath caldera-filling lavas and younger volcanic rocks (33.9–33.4 Ma), but an east-west alignment of paleohills of Proterozoic rocks and parallel exposures of Bonanza and Saguache Creek Tuffs just to the north provide limits to the southern extent of Marshall caldera (Fig. 9). Southeastern margins appear to have been deeply buried beneath younger lavas and clastic rocks of this caldera cycle and the Rawley volcanic complex, then truncated by Bonanza caldera. Based on these relations, Marshall caldera is estimated to be ∼15 × 20 km, with a subsided area of ∼250 km².

No postcollapse resurgence is evident at Marshall caldera, and much of its structure is concealed beneath lavas that overflowed the collapse area. The caldera is exposed at relatively shallow levels, revealing only upper parts of the intracaldera ignimbrite, overlain by postcollapse lavas. Exposed landslide megabreccia, in upper Marshall Creek, is high in the intracaldera ignimbrite, directly overlain by caldera-fill andesite. This geometry suggests abrupt subsidence late during the eruption, after sustained slower subsidence during accumulation of the thick ignimbrite exposed at The Gate, which contains abundant lithic fragments as much as 0.5 m across but lacks discrete interleaved megabreccia lenses.

The postcollapse lavas merge to the south and west with petrologically indistinguishable andesite and dacite of the Rawley volcanic complex (as old as 33.7 Ma) that are precursor to the Bonanza caldera cycle. The postcollapse lavas of Marshall caldera also interfinger with lake-bed deposits of the Pitch-Pinnacle Formation (Olson, 1983; Gregory and McIntosh, 1996). This sedimentary assemblage is thickest beyond the northwest margin of Marshall caldera, in an embayment that appears to have been a broad paleovalley disrupted by caldera-forming events and then occupied by a lake concurrently with eruption of caldera-filling lava flows.

Small remnants of partly welded dacite ignimbrite, interpreted as proximal Thorn Ranch Tuff, on the basis of ages and compositions (Tables 3 and 4; Supplemental Tables 1 and 4) similar to the intracaldera accumulation and widespread outflow east of the Rio Grande rift zone, are preserved locally east and northeast of Marshall caldera. Perplexingly, no outflow remnants of Thorn Ranch Tuff or any other ignimbrite sheet that could be related to Marshall caldera have been identified farther west or south in the San Juan Mountains, even though basal volcanic strata are exposed widely in contact with underlying Mesozoic and Proterozoic rocks (Lipman, 2012). Because of these limited exposures, the eruptive volume of the Thorn Ranch Tuff cannot be determined in detail, but the large area of Marshall caldera (>250 km²), the presence of intracaldera tuff more than 400 m thick with no base exposed, and preservation of outflow tuff at least 70 km east of the source caldera (Fig. 10; McIntosh and Chapin, 2004, Fig. 2D) indicate a major ignimbrite eruption, estimated as 250–500 km³.

Marshall caldera is eroded less deeply than Bonanza, lacks exposed comagmatic intrusions, and exposes only upper portions of the caldera-filling ignimbrite. This geometry is consistent with Rio Grande rift-related regional westward tilting (∼10°–20°) of a large structural block that includes the Sawatch Range and continues south nearly to the valley of Saguache Creek (Fig. 1).

Bonanza caldera, source of the 33.12-Ma Bonanza Tuff (∼1000 km³), is the southernmost and youngest of the ignimbrite centers aligned along the Sawatch Range trend (Fig. 1). In contrast to the multicyclic nested caldera loci for successive eruptions at many large-volume ignimbrite flare-ups elsewhere, Bonanza is an areally isolated collapse structure that formed in response to a single ignimbrite eruption. Although the western topographic rim of Bonanza impinges on Marshall caldera, projected structural boundaries of two calderas appear to be largely or entirely separate.

Existence of a caldera in the Bonanza area was first inferred from a local gravity low (Karig, 1965), followed by recognition that the Bonanza latite of Patton (1916) and Burbank (1932) was welded tuff likely erupted from the Bonanza area (Bruns et al., 1971). As an outgrowth of exploration for porphyry deposits at depth below the mineralized veins at Bonanza, Varga and Smith (1984) synthesized available regional data, augmented by new chemical and isotopic analyses, as evidence for a trap-door caldera of relatively modest size (∼8 × 12 km), in which the subsided block was tilted westward and hinged on its east side. Details of this interpretation are problematic, however, because lavas that overlie the intracaldera ignimbrite are tilted steeply westward along with the underlying ignimbrite. Such a relationship would require the caldera to have subsided after emplacement of the lavas, rather than accompanying the ignimbrite eruption.

Bonanza caldera is here reinterpreted as a much larger resurgent caldera, ∼20 × 25 km across, in which the caldera floor was uplifted steeply after subsidence (Figs. 3 and 9). The previously inferred eastern hinge is now inferred to be the crest of a large elliptical resurgent uplift (Whale Hill dome) within a caldera block that occupies much of the Bonanza map area. Small remnants of east-dipping Bonanza Tuff are preserved on deeply eroded eastern flanks of the Whale Hill dome, but the east margin of the caldera lies concealed beneath the San Luis Valley segment of the Rio Grande rift zone. As a result of the larger caldera size and revised correlation of the outflow ignimbrite, the combined intracaldera and outflow volume of Bonanza Tuff is estimated at ∼1000 km³ (Table 1), in contrast to the estimate of less than 50 km³ by Varga and Smith (1984).

The composite volcanic highland in which the caldera formed is dominated by intermediate-composition to silicic lavas from several eruptive centers (Table 4; Supplemental Table 4). Silicic flows (dacite as well as rhyolite) high in the precaldera assemblage are more abundant at Bonanza than at most other sites of ignimbrite eruptions in the San Juan region and appear to record a prolonged history (33.7–33.2 Ma) of magma assembly and evolution toward evolved compositions. Volcaniclastic deposits interlayered with the lava sequence are less abundant than in similar younger assemblages in the San Juan Mountains, consistent with the inference of a precaldera constructional highland.

Presence of a central highland at Bonanza, prior to eruption of the Bonanza Tuff, is confirmed by deposition of the ignimbrite in broad radial paleovalleys, as especially preserved on the west and south flanks of the caldera (Varga and Smith, 1984). Effects of any precursor tumescence in the Bonanza area are difficult to evaluate but would have been subordinate to constructional processes in generating the overall present-day geometry of the precaldera volcanic sequence. Much of the highland constitutes fill of the earlier Marshall caldera, and many dated lavas are closer in age (33.7–33.5 Ma) to that caldera than to Bonanza. Only on Sargents Mesa and along the south flank of Bonanza caldera, near Saguache Peak, do relatively low-precision whole-rock dates on andesitic lavas (ca. 33.1–33.3 Ma) approach the age of the Bonanza cycle.

In contrast to the one-sided easterly distribution of preserved Thorn Ranch Tuff in relation to its source from Marshall caldera, the Bonanza Tuff is preserved widely both to the east and west of its caldera (Fig. 10). Any deposition of this ignimbrite in the vicinity of the Sawatch Range to the north has been completely eroded, however, and it has been covered by younger volcanic deposits or eroded south and west of the Saguache paleovalley.

The Bonanza Tuff is characterized by large and complex lateral and vertical variations in chemical and phenocryst compositions (Table 4; Supplemental Table 4), from dark crystal-rich silicic andesite and dacite (25%–35% plagioclase, biotite, augite; 60%–66% SiO₂) to light-colored, crystal-poor rhyolite that consists mainly of fine ash (∼5% sanidine, plagioclase, biotite; 73%–76% SiO₂). A distinctive feature of the dacitic tuff, in both intracaldera and outflow deposits, is the presence of several percent of small angular lithics (typically several cm or less), dominantly fragments of andesitic lava. Flattened pumice fiamme are conspicuous in the dacite, commonly 5 cm or longer (Figs. 11B and 11C). In contrast, lithic fragments are rare in the outflow rhyolite, and fiamme are small and obscure. Uniquely among rocks of the SRMVF, even mafic dacite (62%–64% SiO₂) has phenocrysts of sanidine, while some rhyolite contains trace hornblende and titanite.

Prior studies had distinguished a lower dacite (latite) and upper rhyolite, starting with Burbank (1932), but the eruptive significance of the compositional change remained uncertain. From work on the southwest flank of the Bonanza center, Bruns et al. (1971, p. 187) concluded that the “Bonanza Tuff consists of two cooling units” but did not describe the contact. Varga and Smith (1984) later described the Bonanza Tuff as consisting of two sheets, each a single cooling unit, not compositionally comagmatic, and erupted from geographically separate vents. Neither of these studies specifically discussed the presence or absence of evidence for a time break, such as bedded tuff or other sedimentary units, between the two Bonanza cooling units, but the implied interpretation was the existence of two separate ignimbrites, an interpretation that has been continued by others (e.g., McIntosh and Chapin, 2004). In contrast, our work shows that all compositional variations in both intracaldera and outflow Bonanza Tuff occur within a single ignimbrite sheet characterized by compound welding zonations.

West and south of Bonanza caldera, the outflow ignimbrite sheet is dominantly dark- brown crystal-rich dacite, but some proximal sections contain gradational transitions from densely welded lower crystal-poor rhyolite, a thick central zone of dacite, somewhat less welded upper rhyolite, and, locally, an upper welded dacite (Figs. 11 and 12). The compositional transitions correspond imperfectly with welding breaks. For example, the change from main dacite to upper rhyolite is associated with a bench-forming zone of weaker welding between dual cliffs of densely welded tuff, where well exposed along lower Saguache Creek (e.g., Lipman et al., 2013, Stop 1-7), but the upward gradational decrease in crystal content and increase in silica mainly occurs high in the lower cliff rather than coincident with the zone of least welding. Rare partings defined by abrupt changes in pumice or lithic size or abundance within the ignimbrite (Fig. 11A) are interpreted as recording pulsation of eruption intensity or ignimbrite-emplacement dynamics during essentially continuous ignimbrite deposition. No evidence was observed for existence of significant time breaks during deposition, such as interlayered bedded-surge, tephra-fall, or fluviatile ash deposits in either the outflow ignimbrite or in the thick intracaldera accumulation.

In contrast, outflow Bonanza Tuff, which is preserved 80 km beyond the caldera to the east (Fig. 10), was described until recently as a separate rhyolitic ignimbrite, the Gribbles Park Tuff (Epis and Chapin, 1974; McIntosh and Chapin, 2004). The eastern outflow sheet consists dominantly of light-colored, crystal-poor rhyolite but contains a thin internal zone of red-brown dacite, at least at the well-exposed sequence at Two Creek (38°36.6′N, 105°46.7′W), which has been used as a reference section (Epis and Chapin, 1974; McIntosh and Chapin, 2004; Supplemental Tables 1 and 4). Thus, the eastern outflow ignimbrite has a similar general gradational compositional zonation as to the west, but with a much higher proportion of rhyolite to dacite. Total volume of the eastern outflow must be large, at least several hundred cubic kilometers. This unit is not simply correlative with the upper rhyolite zone in proximal western sections as proposed by McIntosh and Chapin (2004, p. 230), who recommended “that the name Bonanza Tuff be restricted to the lower crystal-rich dacitic unit, and that Gribbles Park Tuff be used for the upper rhyolitic unit.” However, our study indicates that all the compositional zones in the outflow ignimbrite, both to west and east, are within a single variably welded ignimbrite sheet, displaying compound cooling (Smith, 1960) but without depositional breaks, for which the name Bonanza has precedence (Patton, 1916; Burbank, 1932). Accordingly, Gribbles Park Tuff is not used as a stratigraphic name.

The thick intracaldera Bonanza Tuff contains even more complex compositional and welding variations within a single ignimbrite unit (Fig. 11). Where a complete ignimbrite section (2.5 km thick) from caldera-floor rocks to postcollapse lavas is exposed on the west flank of the resurgent dome (Fig. 13), rhyolite and dacite compositional zones interfinger as many as 13 times, as identified by variations in crystal content and groundmass color; dacitic tuff is volumetrically dominant, but the eruptive sequence began and ended with rhyolite. The incomplete caldera-fill section on the steeply dipping northeast flank also contains multiple rhyolite-dacite alternations and is at least 1.5–2 km thick, including thick interleaved landslide-breccia deposits (Fig. 14). Other than contacts against landslide breccias, all compositional and welding variations are gradational. The multiple zones of dacitic tuff are typically densely welded (Figs. 11B and 11C); many of the rhyolite zones appear to be less welded, although degree of welding is more difficult to evaluate because of common absence of large pumice and/or fiamme lenses and widespread intense alteration.

The initially erupted lower rhyolite is relatively thin and small in volume within the caldera, varying greatly in thickness within short distances laterally and probably reflecting pre-eruption volcanic topography. This zone is up to ∼100 m thick locally on parts of the southwest flank of the resurgent dome, but its thickness is highly variable; it is absent in other nearby exposures and almost the entire exposed north flank. The dacitic tuff is typically densely welded (Figs. 11B and 11C); in places where propylitically altered, the intracaldera dacite is identifiable as welded tuff only by the typical presence of abundant small angular fragments of andesitic lava. Much of the rhyolite is only moderately welded, but some deep intracaldera zones are densely welded to fluidal, and locally rheomorphic. The lower rhyolite is especially fluidal and lava-like on the west caldera rim (Figs. 11D and 11E) and along the steeply dipping northeast flank of the resurgent dome. Pumice lenses, where distinguishable in fluidal and rheomorphic zones (both rhyolitic and dacitic), commonly define a prominent lineation with prolate elongation ratios of as much as 20:1. Lineations are especially well developed low in the thick ignimbrite section that is so well preserved on the west flank of the resurgent dome; there, they typically plunge nearly directly down dip.

The weighted-mean age for sanidine from intracaldera Bonanza Tuff (33.05 ± 0.06 Ma) is marginally younger than that for the outflow ignimbrite 33.12 ± 0.03 Ma). Although within analytical uncertainty, this difference might reflect prolonged cooling of the multi-kilometer-thick intracaldera accumulation or possible effects of later alteration.

After the ignimbrite eruption, compositionally diverse lavas ranging from andesite to high-silica rhyolite (Table 4; Supplemental Table 4) filled the caldera to overflowing and spread across adjacent slopes to the northwest and southwest. Postcollapse lavas are not preserved on the more eroded eastern side of the caldera, but a distinctive thick lava sequence of crystal-poor silicic dacite (69%–70% SiO₂) on Hayden Peak (Fig. 9) appears to have accumulated in a paleovalley carved deeply through intracaldera Bonanza Tuff. Ages of the upper lava sequence of the Bonanza center, including caldera-filling flows, are bracketed at 33.12–32.25 Ma by the underlying Bonanza Tuff and overlying Saguache Creek Tuff, but the ∼1-km-thick intracaldera sequence capped by Porphyry Peak Rhyolite on the west flank of the Whale Hill dome was emplaced rapidly, by 33.03 Ma (mean of five samples; Supplemental Table 1). The lowermost postcollapse lavas (Squirrel Gulch Andesite) directly overlie upper-rhyolitic ignimbrite of the Bonanza Tuff and were tilted steeply along with the underlying tuff during later resurgent uplift. In contrast, some late caldera-filling lavas (Porphyry Peak Rhyolite and sanidine-bearing dacite) may be tilted somewhat less, as based on contact geometry (primary depositional attitudes are difficult to constrain precisely in these massive viscous flows that commonly form widespread talus on steep vegetated slopes). The tilted intracaldera lavas at Bonanza, erupted prior to and during resurgence, thus are analogous to voluminous early postcaldera rhyolites at Long Valley (∼100 km³, as much as 600 m thick), also erupted during resurgence within ∼100 k.y. after caldera collapse (Bailey et al., 1976).

In contrast to many other large ignimbrite calderas, only small deposits of lake sediments or small late pyroclastic eruptions are preserved in the postcollapse fill at Bonanza. The rarity of intracaldera volcaniclastic rocks at Bonanza is consistent with rapid accumulation of the lavas, largely or entirely prior to major resurgent uplift. The time span between the Bonanza ignimbrite eruption at 33.12 ± 0.03 Ma and that of Porphyry Peak Rhyolite margin that overflowed the north caldera margin by 33.03 ± 0.04 Ma indicates that at least 1 km of lava accumulated in the northwestern caldera area within ∼90 k.y. (as little as 20 k.y. and no more than 160 k.y., based on the 2-sigma uncertainties for the weighted-mean ages).

Compositionally and texturally diverse intrusions varying from gabbroic to silicic granitoid rocks intruded the caldera floor and lowest ignimbrite fill, forming widely scattered exposures that are inferred to represent an irregular roof zone of a more continuous composite body at slightly greater depth. The largest intrusions crop out on the deeply eroded eastern side of the Whale Hill resurgent dome. Texturally diverse areas of granodiorite (56%–62% SiO₂) and intergradational finer-grained phases, which form the 3 × 7 km exposed area of the Turquoise Mine intrusion on the eastern side of the dome (Fig. 9), are compositionally similar to postcollapse andesite and dacite preserved on the western flank. Fine-grained intrusive phases, covering areas as much as several hundred meters across, form bold outcrops of dense dark andesite (55%–56% SiO₂), characterized by rectilinear jointing unlike the hackly fractures that characterize most andesite lavas in the Bonanza area. Some larger areas of the finer-grained phases are composite, containing internal contacts between subunits differing in phenocryst abundance, size, or mode. The areal abundance of intrusive andesite, comingled with coarser granodiorite, is interpreted as representing the roof zone of a large intermediate-composition intrusion that would be less heterogeneous at greater depth. Several small exposures of similar granodiorite to andesite crop out farther west, along tributaries of Kerber Creek, and mineral-exploration drilling on Manitou Mountain (Fig. 3) penetrated granitoid rocks at depths of ∼1 km (Cook, 1960; Gordon Gumble, 2006, written commun.).

A pluton of aplitic to porphyritic granite exposed at near-roof levels in an ∼3 × 4 km area of upper Spring Creek is compositionally similar to postcaldera lavas of the Porphyry Peak Rhyolite. Roof zone and margins of the Spring Creek intrusion are aplitic porphyry (74%–77% SiO₂), containing 15% euhedral K-feldspar; deeper interior portion of the exposed intrusion are also porphyritic but have a medium-grained matrix that is modestly less silicic (72%–73% SiO₂). Microperthitic and weakly argillized K-feldspar from the granitoid intrusions has not been datable with precision comparable to volcanic sanidine, but the somewhat varied cooling ages (32.8–33.3 Ma; Table 3; Supplemental Table 1) overlap those from the Bonanza Tuff and postcaldera lavas. These cooling ages, together with field relationships, indicate emplacement of the resurgent plutons shortly after caldera eruption. Numerous other caldera systems display similarly rapid timing (e.g., review by Lipman and Bachmann, 2015).

A roughly concordant sill-like body of uniform dacite (67%–68% SiO₂), the Eagle Gulch Dacite (latite of Burbank, 1932), was intruded between caldera-floor andesitic lavas and basal intracaldera Bonanza Tuff along a northeast-trending belt ∼8 km long from Kerber Creek to the north slope of Elkhorn Peak (Fig. 9). The sill-like shape and uniform texture and composition of the Eagle Gulch Dacite differ from the more diverse granitoid intrusions, which are interpreted as uppermost levels of a vertically extensive composite intrusion coring the Whale Hill resurgent dome.

Despite detailed complexities, overall evolution of Bonanza caldera has the simplifying attribute of involving only a single large ignimbrite eruption and attendant caldera formation. The diverse exposure levels provide special insights concerning timing of caldera subsidence, distribution of subsidence faults, caldera-floor structures, geometry of resurgent uplift, and ignimbrite and subsidence volume estimates.

The asymmetric areal distributions of the outflow rhyolite and dacite tuffs from Bonanza caldera provide key information on paleotopography and timing of initial caldera collapse (Fig. 15). At inception of the eruption, some barrier on the west side, either prevolcanic structural highlands or earlier volcanic constructs, must have impeded ignimbrite flow; early-erupted rhyolitic ash spread mainly to the east. Beginning of caldera collapse late during eruption of the lower rhyolite appears to have disrupted the western barrier, accompanied by increased eruptive draw-down and/or tapping a new sector of a compositionally complex reservoir that led to initial discharge of voluminous dacite. Dacite was then able to spread widely to the west while accumulating thickly within the subsiding caldera, concurrently with intermittent further eruptions of rhyolitic tuff. The asymmetrical distribution of early- versus later-emplaced compositional phases within the outflow of a single ignimbrite sheet appears somewhat analogous to eruptive processes well documented for eruption of the Bishop Tuff from a zoned magma body (Hildreth and Wilson, 2007). Another example of contrasts in composition and distribution between early- and late-erupted ignimbrite is the Lunar Cuesta Tuff in central Nevada (Best et al., 2013b, fig. 59 and text). The Nevada ignimbrite has a simple reverse compositional zonation, however, in contrast to the multiple compositional oscillations in the Bonanza Tuff.

The relatively modest thicknesses and limited areal extent of the lower-rhyolite zone within the Bonanza caldera provide the primary documentation that caldera collapse began relatively late during this phase of the ignimbrite eruption. If subsidence had accompanied inception of the eruption, or even triggered initial magma expulsion as proposed in some models for large ignimbrite calderas (e.g., Sparks et al., 1985; Lindsay et al., 2001; Gudmundsson, 2008; Gregg et al., 2012; Cashman and Giordano, 2014), a much greater thickness and volume of the early rhyolitic ignimbrite should have ponded within Bonanza caldera. If a volume of early rhyolite tuff comparable to that in the eastern outflow sheet (estimated at 200–300 km³) had accumulated concurrently within Bonanza caldera, the thickness of intracaldera early rhyolite would have been greater than 1 km. Even though later stages of the Bonanza eruption were accompanied by concurrent caldera subsidence, the outflow volume of lower rhyolite alone is comparable to that of several uniform dacite ignimbrites in the SRMVF (Table 1) and elsewhere, suggesting that no simple correlations exist between eruptive volumes or magma compositions and inception of caldera subsidence.

Interfingered with the alternating zones of rhyolite and dacite ignimbrite within Bonanza caldera are many irregular lenses of brecciated precaldera rocks (Figs. 9, 13, and 14), both mesobreccia (Fig. 16) and much larger masses of little-broken massive lava, which are interpreted as landslide debris derived from caldera walls that had become oversteepened during subsidence. Individual blocks in some lenses are larger than outcrops and are termed megabreccia (Lipman, 1976b). Some breccias are heterolithologic on outcrop scale, but in other large areas, blocks are compositionally uniform lava. The most voluminous breccia, locally as much as several hundred meters thick, is low in the caldera fill, close to or in direct contact with caldera-floor rocks (Figs. 13 and 14); boundaries between breccia and floor can be obscure. The deep breccia is best developed along the southwest and north margins of the central resurgent uplift, in proximity to caldera ring faults and the inner wall. Exposures are especially good on the dry southwest-facing slopes of Kerber Creek valley. Much less breccia appears to have reached central areas of the caldera floor, as exposed along the crest of the resurgent dome, and the landslides appear to have thinned with distance from the inner caldera walls. The voluminous deep breccias, in places deposited directly on caldera-floor rocks, are interpreted to record catastrophic initial caldera collapse during the later stages of eruption of the lower rhyolite zone. Thus, intracaldera landslide breccia at the base of an intracaldera ignimbrite sequence need not necessarily document caldera collapse (or vent enlargement) concurrently with eruption inception.

Smaller lenses of meso- and megabreccia interfinger at higher horizons of the intracaldera Bonanza Tuff (Figs. 13 and 17A), indicating that caldera walls became oversteepened intermittently during subsidence, but not as severely as during the initial collapse. Most breccia consists of andesite and dacite fragments from the Rawley complex, but Proterozoic debris is also present, especially within northern sectors of the caldera fill. A Proterozoic source for landslides along this sector would have required deep early subsidence (>1 km), cutting down below the volcanic fill of Marshall caldera. Another major subsidence event late during the eruption is recorded by the ∼1000-m difference in elevation between high remnants of Bonanza Tuff, plastered against the western caldera wall, between Antora Peak and Windy Point (Fig. 5) and the uppermost intracaldera tuff within the caldera structural block that subsided along the Kerber Creek ring fault (Fig. 17A).

Varied thickness and lateral extent among the multiple interfingering zones of rhyolite and dacite in the intracaldera ignimbrite probably reflect diverse factors, including surface irregularities on the pre-eruption lava assemblage, depositional slopes generated by caldera-wall landslide deposits, and mildly asymmetric caldera subsidence. Caldera-floor morphology probably mainly affected distribution and thickness of the lower rhyolite zone. The varied thickness and extent of many compositional zones higher in the intracaldera accumulation are spatially unrelated to breccia lenses. These variations suggest that the ignimbrite depositional surface became weakly tilted in varied directions during the course of caldera subsidence, and tuff accumulated more thickly down slope. Such tilting and asymmetrical collapse have been documented at other ignimbrite calderas (Carr and Quinlivan, 1968; Lipman, 1984, 1997; Branney and Kokelaar, 1994) and in analogue models, especially for early downsagging and trap-door subsidence in small-volume eruptions (Cole et al., 2005; Acocella, 2007). Weakly asymmetrical subsidence at Bonanza is also suggested by rheomorphic structures in rhyolitic tuff deep in the caldera fill and by strongly prolate compaction and elongation of large pumice lenses in dacitic tuff (Fig. 11B). Such flowage structures seem likely in calc-alkaline ignimbrites only when deposited on a slope (e.g., Chapin and Lowell, 1979; John et al., 2008), especially in a dynamic environment of increasing steepness as could occur during caldera collapse.

Despite this evidence for modest asymmetry at times during subsidence at Bonanza, the overall geometry is coherent subsidence of a structural block ∼15 × 20 km across, accommodated along peripheral ring faults. Dips of foliation defined by pumice fiamme (most 35°–55°) do not vary significantly or systematically upward through the 2.5-km section of the intracaldera ignimbrite section exposed on the west flank of the resurgent dome (Fig. 13), indicating that the overall subsidence did not involve sustained progressive tilting. No major fault offsets have been recognized within caldera-floor lavas of the main subsided block; subsidence was dominantly piston style, not piecemeal.

Caldera collapse during eruption of the Bonanza Tuff was primarily accommodated along ring faults that are largely concealed beneath surficial deposits along Kerber Creek and its tributaries (Fig. 9). Fault strands up to several kilometers outboard of the major ring faults accommodated additional subsidence, slumping of large blocks along the south and west caldera margins, and modest inward rotation. As much as ∼4 km of stratigraphic offset occurs along the concealed ring faults on the west side of the caldera, as constrained by a cross section from Whale Hill to Flagstaff Mountain (Fig. 17A). Net offset along the Kerber Creek fault diminishes farther to the south, approaching zero at the junction with Little Kerber Creek and the intersection with the anticlinal crest of the Whale Creek dome (Figs. 3 and 9); the decreased displacement along this southern fault segment is interpreted to result from uplift during resurgence. To the north, caldera faults are largely concealed beneath collapse-related megabreccia and later lavas of the Bonanza eruptive cycle; it remains unclear whether subsidence was as deep as to the south and west. The few other SRMVF calderas that are sufficiently deeply eroded to expose bounding ring faults display varied structural patterns. The multiple interconnected fault strands associated with the single ignimbrite eruption at Bonanza (Fig. 9) differ from the predominantly single ring-fault strand that is well exposed at Lake City, or from the nested faults at Grizzly Peak and probably at Mount Aetna (Lipman, 1975; Shannon, 1988; Fridrich et al., 1991).

Andesitic and dacitic lavas of the caldera floor that are adjacent to ring faults along Kerber Creek and to the south are locally severely shattered, involving textures and structures that seemingly have not been widely recognized at calderas elsewhere. Angular blocks mostly less than 0.5 m across are juxtaposed, with only minor matrix of comminuted lava (Fig. 18). In many exposures, finely shattered fragments fit together without large-scale rotation or other movement, and such rocks grade into more massive lavas of the caldera floor within 100–200 m away from mapped ring faults. In some zones, breccias with angular and rounded clasts are matrix supported, but despite areal proximity to megabreccia at the base of the intracaldera Bonanza Tuff, no tuffaceous component is present in the shatter breccia. Although fault planes, slickensides, or other evidence of offset are sparse, the close proximity of the shatter breccias to the main caldera-collapse faults suggests that they formed during subsidence (or resurgence), perhaps due in part to hydraulic fracturing. Temperatures must have remained low during fracturing, however, as evidence is absent for silicification or other mineral precipitation from high-temperature hydrothermal fluids such as described along some faults elsewhere (Caine et al., 2010, and references therein). Compositions of breccia matrix differ little from bulk andesite compositions, other than modestly variable alkali ratios and higher loss on ignition values (Supplemental Table 4). Alternatively, and perhaps more likely, shattering of the floor rocks may have resulted from compression and crushing of the subsiding structural caldera block in proximity to steeply inward-dipping ring faults. The shatter breccias at Bonanza somewhat resemble the “collar breccia” at Indian Peak caldera, southern Nevada-Utah (Best et al., 2013a, especially their fig. 40), interpreted by these authors as having formed by landsliding and ring-fault–related fracturing.

Also interpreted as related to caldera subsidence are several east-trending arcuate faults south of Kerber Creek, which juxtapose Paleozoic sedimentary formations against Proterozoic rocks (Fig. 9). These were described by Burbank (1932, p. 39–40, Plate 3) as low-angle thrust faults of prevolcanic age, but no specific evidence was cited for a thrust interpretation, other than the presence of Proterozoic granite on high ridges south of the southward-dipping Paleozoic strata exposed low along Kerber Creek. Detailed tracing of faults across ridges and gullies, with better base-map and geographic positioning system (GPS) control than was possible for Burbank, documents generally steep dips, at least 45°–60°, although vertical relief is insufficient to quantify fault dip precisely. Accordingly, rather than low-angle thrusts active during Laramide compression, these faults are here interpreted as large-scale block slumps of mid-Tertiary age, related to peripheral subsidence and failure along oversteepened walls along the south margin of Bonanza caldera (Fig. 17B, line 2). Rotation during slumping could have produced southward dips of the Paleozoic strata that crop out at low elevations. Additionally, at least some component of the southward dips in the prevolcanic rocks likely resulted from tilting along the lower south flank of the large resurgent dome within the caldera.

Particularly revealing within Bonanza caldera are the areally widespread exposures of structurally coherent caldera-floor lavas and basal deposits associated with the ignimbrite eruption and caldera collapse. Other than the thick dipping section of Bonanza Tuff on the lower west flank, virtually the entire resurgently domed caldera floor has been erosionally exhumed at stratigraphic levels close to original contacts with the basal intracaldera ignimbrite, over an area ∼10 × 15 km across constituting much of the ring-fault–bounded caldera structural block (Fig. 9).

At Bonanza, stratigraphic levels close to the original caldera floor are exposed for about two-thirds of the area of the ring-fault–bounded structural core (the little-broken thick sequence of overlying intracaldera ignimbrite covers ∼15% of the west flank of the resurgent dome, while on the eastern side, surficial deposits of the San Luis Valley conceal an additional 15%–20%). Exposures of the stratigraphic transition from caldera-floor lavas to ignimbrite fill at Bonanza are especially good on the relatively dry and weakly vegetated south-facing slopes above Kerber Creek. In this area fractured but seemingly coherent thick sequences of andesite and dacite lavas merge imperceptibly upward into shattered outcrops of similar lava, between which irregular crack fills and pockets of dacitic and rhyolitic tuff form matrix between megabreccia blocks. Within the near-floor megabreccia, most good outcrops consist of erosion-resistant intermediate-composition lavas. Much of the matrix tuff is weakly welded, lithic rich, and exposed only as fragments on slopes. As a result, contacts between caldera floor versus caldera-fill megabreccia can be broadly gradational and locatable only approximately in many places.

Such deep levels of caldera structure have rarely been observed on a comparable areal scale elsewhere, where floor rocks are seen mainly in oblique cross sections through structurally disrupted and tilted caldera remnants, as in the Great Basin of the western United States (e.g., Best et al., 2013a, 2013b; Henry and John, 2013). Caldera-floor levels of greater areal extent are well exposed at several well-documented large ignimbrite calderas of Paleozoic age; e.g., Ordovician Scafell and Glen Coe calderas in Great Britain (Branney and Kokelaar, 1994; Moore and Kokelaar, 1998) and Permian Sesia and Ora calderas in northern Italy (Quick et al., 2009; Sbisà, 2010; Willcock et al., 2013), but these differ from Bonanza in important aspects, including morphologic preservation, eruptive history, and tectonic setting. The record of explosive volcanism at all these sites comes mainly from thick intracaldera accumulations that have been deeply eroded; little or no outflow ignimbrite or features of near-surface caldera morphology are preserved. All four Paleozoic calderas are interpreted to have formed in extensional tectonic regimes, where regional faults strongly influenced subsidence geometry. None of the Paleozoic calderas are resurgently domed. Both British Ordovician calderas are polycyclic, each having erupted multiple large ignimbrites, separated by significant time breaks as documented by phreatomagmatic and sedimentary interbeds. Thus, recurrent subsidence was likely, perhaps at differing loci within the overall caldera complex, and the documented piecemeal-style disruption of their floors may have resulted from composite disruption during the multiple collapse events. The thick and texturally spectacular Ora ignimbrite has been interpreted as a single deposit, erupted sequentially from two separate caldera loci that each subsided relatively coherently, but locations and nature of original caldera walls and bounding faults seem widely obscured by postvolcanic regional faults.

Where well exposed, the tuff matrix in much of the caldera-collapse breccia is only weakly welded and distributed as highly irregular seams (Fig. 16), but in some outcrops tuffaceous crack fills have dike-like shapes, are strongly welded, and contain steeply dipping pumice fiamme (Fig. 19). The crack fills are relatively thin (typically <1–2 m), discontinuous (commonly traceable only for a few tens of meter), and irregular in shape and trend. In places, fiamme-rich welded tuff grades along strike into flow-laminated crystal-poor rhyolite that lacks obvious fragmental textures; a few parallel dike-like bodies consist entirely of rhyolite without lithic fragments or other surviving pyroclastic textures.

The best exposed fracture fills of highly welded and rheomorphic rhyolitic tuff, on slopes north of Kerber Creek, tend to be parallel to adjacent caldera ring faults, but all identified fracture fills are located near the transition from caldera floor upward into caldera-fill megabreccia and matrix tuff. No comparable fracture-fill tuff has been found at deeper exposed levels of caldera-floor lavas or in underlying Paleozoic and Proterozoic rocks. The highly welded to fluidal rhyolite in the dike-like fracture fills in places merges with areas of less welded tuffaceous matrix in the megabreccia and with larger pockets and lenses of more uniform lower rhyolite of the intracaldera Bonanza Tuff that dip conformably with the flanks of the resurgent dome. Most fracture fills with identified pyroclastic textures consist of crystal-poor rhyolite (only one small outcrop of dacite fracture-fill tuff was found), similar to that of the early-erupted lower rhyolite phase of the ignimbrite sheet as would be anticipated if caldera subsidence began during this stage of the eruption. Similar local pods of pumiceous to fluidal rhyolite are present elsewhere on flanks of the resurgent dome, especially within large northern areas interpreted as megabreccia, but these areas are more vegetated and exposures are limited. Many of the crack-fill tuffs appear originally to have been glassy, even where most welded and fluidal, as suggested by well-preserved relict pumice and shard textures and by extreme alkali exchange (K₂O/Na₂O ratios, commonly 3-10) compared to analyzed samples of the outflow and intracaldera ignimbrite (typical ratios, 1.3-1.5: Supplemental Table 4).

Some of the ignimbrite crack fills at Bonanza could be interpreted as intrusive dikes of fluidal rhyolite or vent fissures for ignimbrite eruptions, but their stratigraphic distribution, lateral textural variations, and compositions suggest that they are best interpreted as surficial fills between blocks of early caldera-collapse megabreccia, injected down into cracks that opened dilatantly during caldera subsidence. Dike-like pyroclastic bodies at several ignimbrite and caldera settings elsewhere have been similarly interpreted as dilatant crack fills (Lipman, 1964; Branney and Kokelaar, 1994, p. 525; Best et al., 2013a, p. 920). Somewhat similar to the Bonanza crack fills are wider and more laterally continuous pyroclastic intrusions that have been discussed elsewhere as possible ignimbrite vent structures: e.g., a welded-tuff fissure near a caldera margin in central Nevada (Ekren and Byers, 1976), the tuff dike at Mount Aetna caldera in the Swatch Range (Shannon et al., 1987), elongate steep bodies of welded tuff in the Grizzly Peak caldera (Fridrich et al., 1991), pyroclastic dikes at Fairview Peak and Caetano calderas in Nevada (Henry and John, 2013, p. 977, 993), some tuff-filled fissures at the Scafell caldera in the British Lake District (Branney and Kokelaar, 1994, p. 526), or the Big Butte pyroclastic complex in Montana (Houston and Dilles, 2013, p. 1407).

Interpretation of the Bonanza fracture fills as marking primary eruptive sites seems improbable, however, because (1) these structures are localized near the interface between caldera-floor lavas and overlying ignimbrite and megabreccia fill, (2) in places steeply dipping fracture-filling tuff is traceable continuously into the lower rhyolite zone of intracaldera Bonanza Tuff, (3) the fracture fills are relatively small and discontinuous in comparison to pyroclastic dikes described elsewhere, and (4) the predominance of rhyolitic compositions in the fracture fills seems inconsistent with interpretation as primary vents because the peak stages in the Bonanza eruption were dominated by dacitic ignimbrite. For interpretation of the Bonanza crack fills as eruptive sites, the vent geometry would have been an areally-widespread diffuse network of weakly interconnected small fractures, without preservation of sizable discrete eruptive loci along the caldera ring faults.

The well-preserved morphology and structures at Bonanza, especially the exposed caldera floor and segments of the topographic rim, provide unusual opportunities to evaluate proximal precursor activity prior to a major ignimbrite eruption in the SRMVF. Notably, no silicic deposits have been recognized, either lavas or Plinian tephra, that closely preceded the ignimbrite eruption.

Several lavas of crystal-poor, sanidine-bearing rhyolite, high in the caldera-floor lava assemblage along east and south slopes of the resurgent dome and on the proximal south flank of Bonanza caldera, were initially considered potential candidates for ignimbrite precursors because of their similar modal mineralogy to rhyolitic Bonanza Tuff. However, these lavas have modestly different trace-element compositions than the ignimbrite (Supplemental Table 4), their sanidine phenocrysts are more potassic, and ⁴⁰Ar/³⁹Ar ages (ca. 33.7 Ma; four sites) are ∼0.5 m.y. older (Table 3; Supplemental Table 1). Other rhyolitic lavas high in the precaldera lava sequence (Fig. 9) lack sanidine phenocrysts and cannot be dated precisely, but these also have trace-element compositions dissimilar to rhyolitic Bonanza Tuff and appear to be typical of the Conejos-type rhyolites that are distributed sparsely but widely in the northeast San Juan region (e.g., Lipman, 2012). Despite the diverse erosional levels at Bonanza, exposures of the precaldera assemblage are far from complete, and precursor lavas could remain hidden. However, the absence of silicic eruptions of appropriate composition or age suggests that the rhyolitic Bonanza magma was assembled shortly before inception of the ignimbrite eruptions, with few if any lava precursors reaching the surface.

In addition to apparent absence of precursor silicic lavas at Bonanza, no initial tephra-fall deposits of Plinian type are preserved beneath the ignimbrite. The deep dissection of the caldera fill at Bonanza locally provides well-exposed contacts between basal ignimbrite (dacite, as well as rhyolite) and underlying caldera-floor rocks. Both on flanks of the resurgently domed caldera floor and along the western caldera rim, proximal Bonanza Tuff is commonly welded to its base and directly overlies precaldera lavas, without intervening bedded tephra (Fig. 11D).

Similarly, no thick or widespread Plinian fall deposits have been recognized beneath other large ignimbrite sheets in the SRMVF, including those with rhyolitic or zoned compositions. Plinian deposits also appear to be rare or absent in association with compositionally diverse ignimbrite eruptions in other calc-alkaline Cordilleran systems, such as the southern Great Basin ignimbrite field (Best et al., 2013a, 2013b), the central Andes (Sparks et al., 1985; Lindsay et al., 2001; de Silva et al., 2006), or the 74-ka Toba eruption in Indonesia (Chesner, 2012). This contrast with thick Plinian deposits at well-known younger calderas such as Long Valley, Yellowstone, and Taupo raises questions about recent proposals that precursor tephra deposits typically form at inception of rhyolitic ignimbrite eruptions but not during large dacitic eruptions, in response to contrasting mechanisms and timing of caldera subsidence (summarized by Cashman and Giordano, 2014, and references therein). Explosive discharge of hundreds to thousands of cubic kilometers of silicic magma seems improbable without concurrently generating a high thermal column of buoyantly convecting ash at the vent(s), analogous to that in smaller historic eruptions.

Alternatively, the presence or absence of a proximal basal Plinian deposit may be related to timing of initial ignimbrite generation. A rate of magma discharge that was high from inception of the eruption could generate large pyroclastic flows rapidly, such that proximal ash and pumice fall would settle and mix with the moving pyroclastic flow, without forming an underlying fallout layer; Plinian fallout deposits would be preserved only distally beyond the ignimbrite sheet. Such a process was observed on a small scale on 7 August 1980 at Mount St. Helens, where the pyroclastic flow was emplaced while the Plinian column continued to rise, and proximal tephra fell to the ground only after termination of movement in the pyroclastic flow (Christiansen and Peterson, 1981, their fig. 15). Contrasting crystal contents in matrix versus pumice indicate that large volumetric proportions of ash were elutriated during eruption of several well-studied large ignimbrites that lack underlying Plinian-fall deposits; the winnowed fine ash must have been deposited on the moving pyroclastic flow and at more distal sites (e.g., Lipman, 1967; Walker, 1972; Folkes et al., 2011; Chesner, 2012). Such an explanation for the absence of Plinian layers beneath large ignimbrites has also been proposed briefly by Branney and Kokelaar (2002, p. 7 and their fig. 6.5). For the SRMVF ignimbrites that lack associated proximal-fall deposits, distal Plinian tephra may exist among the many ash beds in the Oligocene White River Formation on the High Plains, Colorado and Wyoming (e.g., Prothero, 1996; Larson and Evanoff, 1998), although no systematic studies have thus far attempted detailed correlations. Similarly, as much as several hundred cubic kilometers of fallout ash in the midcontinent has been interpreted as correlative with individual large dacitic ignimbrites of mid-Tertiary age erupted in the southern Great Basin (Best et al., 2013a, 2013b).

After the ignimbrite eruption and emplacement of at least most caldera-filling lava flows, the caldera floor was arched into a spectacularly large and steep-sided resurgent dome that is gently arcuate to the east (Figs. 3, 9, 17A, and 20). Erosion has stripped most intracaldera Bonanza Tuff from the floor lavas along the crest of the dome, which is well defined at present by the gentle upland surface on Whale Hill (Fig. 20). Small remnants of subhorizontal Bonanza Tuff preserved along the crest of the resurgent structure at Round Mountain and Elkhorn Peak provide critical constraints on caldera-floor geometry and structure along highest parts of the dome. Notably, caldera-floor lavas are at elevations above 3700 m on the crest of the resurgent dome, as high as comparable units on the west topographic rim (Fig. 17A). This relation indicates ∼3.5 km resurgent uplift, equal to the subsidence documented by the thick caldera-fill section of ignimbrite and pre-resurgent lavas on the west flank of the dome. If thickness of the caldera fill centrally within the caldera had been similar to the west-flank section, the original dome crest would have had an elevation of ∼7000 m above present-day sea level.

Dips on flanks of the Whale Hill dome are uniquely steep and variable compared to resurgent uplifts at other well-documented ignimbrite calderas: 40°–60° on the west side of the dome where widespread preservation of intracaldera tuff provides robust structural control, nearly vertical along parts of the northeast and southwest flanks (Figs. 13, 14, and 17A), but typically only 20°–30° on the southeast flank. This asymmetry may result in part from variable tilting of the Bonanza region, within a zone of structural transfer between segments of the Rio Grande rift zone. The dome appears to have been largely bounded and partly accommodated by ring faults that initially formed during caldera subsidence. Resurgence is inferred to have been caused by emplacement of multiple intrusions centrally within the caldera, including the granodiorite to granite bodies that are exposed on the eastern flank of the dome.

The south margin of the resurgent dome is especially well constrained by the southeast-plunging Proterozoic-cored anticline and flanking Paleozoic sedimentary strata (Figs. 9 and 17B) that are stratigraphically much more precisely defined than the precaldera lava succession. The southern anticline has been previously interpreted as a prevolcanic (Laramide) structure (Burbank, 1932; Tweto et al., 1976; Cappa and Wallace, 2007). Dips of immediately overlying volcanic strata including Bonanza Tuff, although less widely measurable, locally are nearly as steep on the fold flanks and have a similar asymmetry as the well-stratified Paleozoic sedimentary strata (Fig. 21), however, requiring that much of the tilting was postcaldera. At least the main development of the fold accordingly must be Tertiary in age; it is here interpreted as the southern continuation of the elongate Whale Hill dome. Because the Proterozoic rocks in the core of the anticline are directly overlain by Tertiary lava flows, without intervening Paleozoic sedimentary strata, this area must have also been a prevolcanic high. Perhaps a more open fold in the prevolcanic rocks influenced the location of postcaldera resurgence, or alternatively this area was simply a paleohighland along a northwestern erosional truncation of Paleozoic strata.

Aspects of the resurgent caldera structure at Bonanza were anticipated by Burbank (1932, p. 42–43), who concluded that “arching and tilting of the formations . . . is believed to have been initiated by the intrusion of a large body of molten lava . . . The crust was consequently bulged upward, blocks of it were tilted in different directions.” Prior mapping depicted a highly intricate mosaic of rectilinear faults in the mining district (Burbank, 1932, plate 1), but many of the depicted faults were required to accommodate an overly simplified stratigraphic sequence, without adequate available concepts of ignimbrite-eruption and caldera-filling processes.

In the current study, only a few faults with documentable displacement have been identified confidently within the resurgent dome. Evidence for sizable fault displacements has been elusive on heavily vegetated slopes where talus is widespread, outcrops rare, reliable stratigraphic-marker horizons sparse, and underground mines no longer accessible. More faults than have been mapped are likely within the resurgent block, but major fault repetitions seem unlikely. In contrast to the work by Burbank (1932), Patton (1916, p. 63), with access to more of the underground mine workings in the Bonanza district, noted that “While minor faults involving a movement of a few inches or, at most, a few feet are of common occurrence, no evidence of faulting on a large scale has been discovered.” Beyond the mining district, relatively coherent subsidence and subsequent resurgence of the caldera floor are well documented in the southern caldera area, where detailed structural control is provided by the high-resolution stratigraphy of the regional lower Paleozoic formations. These strata dip steeply on flanks of the resurgent dome but are traceable continuously across the nose of its south-plunging anticlinal termination without sizable fault displacements other than by the major caldera ring faults.

Despite the scarcity of mapped faults on the resurgent dome at Bonanza, some disruption likely accompanied uplift, as at other well-studied resurgent calderas that are characterized by keystone grabens and other uplift-related faults (e.g., Valles, Creede, Timber Mountain, Lake City, and Cerro Galan: Smith and Bailey, 1968; Steven and Ratté, 1973; Byers et al., 1976; Lipman, 1976a; Folkes et al., 2011). The apparently more limited fault disruption of caldera floor and resurgent dome at Bonanza may be related to formation of this caldera in response to a single ignimbrite eruption; the other resurgent calderas just noted are nested within earlier ignimbrite subsidence structures, and prior disruption of the subsided areas may have contributed to more complex fracturing and larger displacements during resurgence. The relatively limited faulting at Bonanza may also partly account for the modest mineralization there, in comparison to otherwise analogous epithermal vein systems in SRMVF caldera settings such as Creede and Silverton.

Uplift of the Whale Hill dome was geologically rapid: 3.5 km at caldera-floor level in less than 100 k.y. (Table 3), as bracketed by ages of the tilted caldera-filling Bonanza Tuff (33.12 Ma) and Porphyry Peak Rhyolite (33.03 Ma). The average resurgence rate (3.5 cm/yr) at Bonanza is roughly similar to that well constrained for uplift of the Samosir Island resurgent dome at Toba caldera for the interval since ca. 34 ka, declining from ∼4.9 cm/yr to <1 cm/yr, with a suggested long-term average of 2–3 cm/yr (de Silva et al., 2015).

Several tantalizing features suggest that resurgence at Bonanza may have been rapid, while deep parts of the caldera-fill ignimbrite remained hot and ductile: (1) extreme fluidal welding and lava-like flowage of the lower rhyolite unit at the base of the thick intracaldera ignimbrite accumulation (Figs. 11D and 11E); (2) down-dip trends of prolate fiamme lineations in fluidly welded ignimbrite on flanks of the dome; (3) apparently limited brittle-fault disruption of caldera-floor levels in the Whale Hill dome, in comparison to complex keystone and other faults in resurgently domed calderas elsewhere; (4) lower dips in upper lavas of the caldera fill than in initially erupted lavas and underlying Bonanza Tuff, suggesting rapid resurgence concurrently with the accumulation of these lavas as documented by isotopic ages; and (5) marginally younger sanidine ages from intracaldera Bonanza Tuff than from outflow portions (Fig. 4), compatible with prolonged slow cooling deep in the intracaldera ignimbrite.

A notable feature of Bonanza is the geologically recent exhumation of the western side of the caldera to morphology closely approximating its primary volcanic features (Fig. 5), while concurrently exposing deep levels on its eastern slopes, including caldera floor, the pre-Bonanza lava assemblage, and underlying Paleozoic sedimentary strata and Precambrian basement rocks (Fig. 9). By late in Oligocene time, the Bonanza caldera area was likely buried to considerable depth by continued local volcanism and by the many younger ignimbrites erupted from farther southwest in the San Juan region. More recently, rift-related tilting and deep erosion into the rugged present-day landforms have exposed the interior of the caldera at levels from Precambrian basement, through the Paleozoic sedimentary sequence, precaldera intermediate-composition lavas, the entire section of intracaldera tuff, overlying postcaldera lavas, and associated resurgent intrusions.

Despite the later depositional, erosional, and structural complexities, parts of the present-day topography have been eroded to features notably similar to those of the Oligocene caldera. These include the high present-day topographic crest of the Whale Hill resurgent dome (Fig. 20), even though almost entirely stripped of intracaldera Bonanza Tuff; the entire ∼3.5 km thickness of caldera-filling tuff and postcollapse lavas on the west flank of the dome (Figs. 13 and 17); western tributaries to Kerber Creek that coincide broadly with an original moat between inner caldera wall and flank of the resurgent uplift (Fig. 9); and the high western ridges of Antora Peak (13,269 ft, 4044 m), Windy Point (11,900 ft, 3627 m), Flagstaff Mountain (12,072 ft, 3680 m), and farther south (Fig. 5) that coincide roughly with the original topographic rim of the caldera. Although more modified by erosion, the continuation of high ridges southeast across Ute Pass to Saguache Peak (10,550 ft, 3215 m) and then eastward (Fig. 3) seems likely to represent further approximations of the caldera rim (Fig. 9). The caldera elements viewed up Kerber Creek valley, from toward Antora Peak (Fig. 5), appear broadly similar to those right after completion of the Bonanza resurgence, except perhaps that the caldera-moat valley would have been largely filled by postcollapse lavas. Thus, the morphology on the western side of Bonanza caldera is broadly comparable to younger San Juan calderas such as Creede and Cochetopa Park, where later Cenozoic erosion has exhumed Oligocene caldera morphology that is even more completely preserved (Steven and Ratté, 1973; Steven and Lipman, 1976; Lipman and McIntosh, 2008).

Large segments of the west caldera rim are still capped by thick Bonanza Tuff, although some of these west-dipping exposures may represent high remnants of caldera-filling ignimbrite, ponded between the ring-fault zone and the caldera-wall “collar” (Lipman, 1997) that had been enlarged by landsliding during caldera collapse. The incomplete section of Bonanza Tuff, with its top eroded, is more than 300 m thick on the northwest flank of Antora Peak; such a great thickness of proximal ignimbrite seems unlikely to have accumulated high on the preexisting edifice of the Rawley volcanic complex prior to caldera collapse. When the geometry of the west-rim ignimbrite remnants is corrected for their 15°–20° westward dips, they appear as westward-thinning scabs, banked against older intermediate-composition lavas of the inner caldera wall. Similar high-fill remnants of intracaldera ignimbrite, preserved as continued subsidence dropped the central caldera floor to greater depths, are preserved at other San Juan calderas such as Lake City and Creede (Lipman, 1976a; Lipman, 2000; Figs. 13 and 14).

In contrast, the eastern Bonanza caldera is exposed at much deeper levels, and accordingly little of the caldera morphology is preserved. Only a few small scabs of intracaldera tuff have survived erosion on the southeast flank of the Whale Hill dome, and erosion has widely cut through the precaldera lavas of the caldera floor down into Paleozoic and Precambrian basement rocks.

The east margin of Bonanza caldera is largely concealed beneath alluvial fill of the San Luis Valley in the northern Rio Grande rift. However, any geometrically simple arcuate projection of the Kerber Creek ring fault and the morphologic topographic rim suggests that the original east caldera margin lay close to the major bounding normal faults at the west base of the present-day Sangre de Cristo Range from south of Poncha Pass to east of Villa Grove (Figs. 3 and 9). In a small erosional patch low along the steep west-facing range front southeast of Poncha Pass (Fig. 9), dacitic Bonanza Tuff and andesitic lavas lap onto Proterozoic metamorphic rocks (Van Alstine, 1975), perhaps defining a distal wedge of intracaldera ignimbrite lapping out against the eastern caldera wall. Because the Bonanza eruption preceded formation of the rift zone and uplift of the Sangre de Cristos, thus permitting the ignimbrite to spread widely to the east (Fig. 10), this patch of Bonanza Tuff at low elevation along the mountain front would have been much deeper prior to rift formation and uplift of the Sangre de Cristo block. Alternatively, these small exposures might be interpretable as preserving proximal outflow of the Bonanza Tuff, if future studies were to document major rift faults farther east, higher within exposed Precambrian rocks on the west-facing slope of the Sangre de Cristo Range.

Combining remnants of the original topographic rim on west and south sides of Bonanza with projection across the San Luis Valley to the inferred east wall against the northern Sangre de Cristo Range defines an approximate overall topographic caldera ∼20 × 25 km across. This topographic caldera would have been larger than the structurally subsided floor, because of landsliding and block slumping from oversteepened inner walls during the continued ignimbrite eruption (Lipman, 1997). The arcuate valley of Kerber Creek, underlain by the main arcuate ring-fractures that accommodated caldera subsidence (Fig. 5), appears to follow the original morphologic moat between the inner caldera wall that had been enlarged by landsliding and the south and west flanks of the resurgent dome. To the north and northwest, the structural boundary, and even the topographic rim, are largely obscured by fill of postcollapse lava flows, but the presence of thick megabreccia in this sector suggests deep subsidence and proximity to a structural boundary at greater depth. Projection from the better constrained south and west sides suggests that the ring faults bound an elliptical structural caldera ∼15 × 20 km across. About two-thirds of this structural area is exposed at present-day levels of caldera floor; ∼20% is covered by intracaldera ignimbrite and overlying lavas, mainly on the west side of the resurgent dome, and ∼15% is concealed beneath sediments in the San Luis Valley.

The resulting three-dimensional preservation of morphologic, stratigraphic, and structural caldera features at Bonanza includes shallow postcollapse fill and only modestly modified segments of the morphologic topographic rim and inner caldera walls, down through the entire intracaldera-filling ignimbrite and megabreccia, to ring faults, resurgent intrusions, and prevolcanic basement, with resulting constraints on eruptive and structural evolution during a single caldera cycle. Such extensive exposure of these diverse caldera elements is unique in the Southern Rocky Mountain volcanic field and perhaps among large Cordilleran-arc calderas associated with ignimbrite flareups elsewhere. The beautifully preserved young calderas of the Altiplano volcanic complex and Cerro Galan in arid high regions of the South American Andes lack exposure of deep levels of intracaldera fill and floor rocks or any associated subvolcanic intrusions. In the semi-arid Great Basin of the western United States, tilting by younger faults provides exceptional oblique exposures of mid-Tertiary ignimbrite caldera fragments and their intracaldera deposits while preserving widespread outflow equivalents, but overall caldera geometry is obscured by the extensional disruption and sediment-filled basins between dismembered volcanic sections. Deeply dissected Ordovician calderas (Scafell, Snowdonia, and Glen Coe) in Great Britain display structurally complex caldera floors and ignimbrite fill, but features directly associated with individual eruptions are difficult to separate at these multicyclic ignimbrite centers, and the associated outflow ignimbrites have been completely eroded (Branney and Kokelaar, 1994; Moore and Kokelaar, 1998).

As elsewhere in the SRMVF, estimates of the magmatic volume erupted as ignimbrite are highly approximate at Bonanza, because of incomplete exposure, variable thickness due to deposition on irregular paleotopography, and extensive subsequent erosion. Total eruptive volume of the Bonanza Tuff is estimated at ∼1000 km³ (Table 1), based on intracaldera and outflow distribution and thickness. The volume of intracaldera ignimbrite is estimated to be greater than 500 km³, based an area of 250 km² of the subsided structural block (∼15 × 20 km across), and an average tuff thickness of 2 km or greater. While much of the intracaldera tuff has been eroded, its thickness is as much as 2.5 km in the complete section on the west flank of the resurgent dome and more than 1.5 km in the steeply dipping incomplete section on the northeast side (Figs. 13, 14, and 17A). Additional ignimbrite fill that would have accumulated in the “collar” area between the inner topographic wall and the ring fault, but its volume would have been partly counterbalanced by the caldera-wall landslide deposits within the structural block. The preserved extent of outflow Bonanza Tuff, still densely welded at least 40 km west from caldera rim (Lipman, 2012) and 70 km to the east (Fig. 10; McIntosh and Chapin, 2004) and typically several tens to as much as a hundred meters thick, yield an estimated volume of several hundred cubic kilometers for each sector. Thus, even though deposition along paleovalleys, erosional removal of original deposits to the north, and cover by younger volcanic rocks to the southwest in the San Juan region preclude more detailed estimates for Bonanza, intracaldera and outflow volumes thus seem roughly subequal, a crude approximation that appears to hold for many large ignimbrites (Lipman, 1984; Mason et al., 2004; Folkes et al., 2011).

In comparison, the volume of total structural subsidence at Bonanza caldera is also estimated at 750–1000 km³, based on average vertical subsidence of >3 km for the relatively coherent equant block within the ring fault (>2 km of intracaldera ignimbrite, overlain by >1 km of caldera-filling lava). Comparable multi-kilometer subsidence is also documented by minimum thickness of caldera-filling ignimbrite and postsubsidence lavas at other large mid-Tertiary calderas such as Lake City and Grizzly Peak in the SRMVF (Lipman, 1975, 2007; Fridrich et al., 1991) and elsewhere in the U.S. Cordillera (Henry and John, 2013; Best et al., 2013a, 2013b), but the extent of exposed caldera floor at Bonanza seems unique.

A corollary implication of the high-standing caldera floor on the Bonanza resurgent dome is that intracaldera ignimbrite volume could be overestimated and potentially yield overly large ratios of intracaldera versus outflow accumulations, if based on exposed maximum topographic relief of the resurgent uplift as assumed for some little-eroded young calderas (e.g., Lindsay et al., 2001; Salisbury et al., 2011). For Bonanza, ring-fault locations, subsidence depth (∼3.5 km, from thickness of intracaldera ignimbrite and postcollapse lava fill), and similar height of resurgent uplift are relatively well constrained. These then define proportions of intracaldera ignimbrite versus uplifted caldera floor in the resurgent structure. Based on a simple ellipsoidal model for uplift at Bonanza, caldera-floor rocks form about one-third of total dome volume. Somewhat analogously, intracaldera volume of the 2.1-Ma climactic ignimbrite at the superbly preserved Cerro Galan caldera in the central Andes is now estimated at 315 km³, reduced from an earlier estimate of 500 km³ (Sparks et al., 1985) largely due to recognition of high-standing basement rocks on the southern flank of the dome (Folkes et al., 2011).

In contrast, volumes of outflow ignimbrites may tend to be underestimated because of common rapid erosion, especially for weakly welded deposits. The Thorn Ranch Tuff, as well as the three large early Oligocene ignimbrites erupted from Sawatch Range calderas (Wall Mountain, Grizzly Peak, and Badger Creek Tuffs), are nearly or completely absent west of their source calderas in comparison to widely exposed areas to the east (McIntosh and Chapin, 2004). The few small western remnants of Wall Mountain Tuff, even though ∼70 km southwest of its inferred source above Mount Princeton, are still thick (to >100 m) and rheomorphically welded (Lipman, 2012); accordingly, this ignimbrite originally would likely have been about as voluminous and widespread to the west as its much more widely preserved eastern distribution. This present-day asymmetry suggests that erosion was rapid along higher slopes of Laramide-age Rocky Mountain uplifts, soon after ignimbrite eruptions from sources in the Sawatch Range. In comparison, ignimbrites erupted from farther west and south in the San Juan region, where the volcanic rocks were deposited mainly on less deformed strata of the northeastern Colorado Plateau, tend to be preserved widely in all directions around their caldera sources (Steven and Lipman, 1976). Even there, however, weakly welded upper zones of the ignimbrite sheets are commonly eroded, down to dense interiors, and eruptive volume may be underestimated. In an impressively detailed evaluation of magmatic volumes for the young Cerro Galan ignimbrite, Folkes et al. (2011) estimated that the exposed outflow deposit (108 km³) constitutes only about a quarter of the original outflow volume (486 km³). Thus, much of the deposit has been lost in only 2.1 m.y., even in a highly arid area where the dominant erosion has been by wind.

Because of such uncertainties, inferences about ratios between intracaldera and outflow ignimbrite volumes should be viewed only as rough approximations (e.g., discussion in Mason et al., 2004, p. 737), to be applied with caution for interpretations such as time of initial caldera formation. Inception of collapse appears to correlate roughly with eruption size; small calderas collapsing late during the eruption, larger ones forming earlier (reviewed by Cashman and Giordano, 2014). Initial subsidence at some large calderas has even been inferred to precede and trigger eruption of voluminous dacitic ignimbrites from low fountains without generating fallout deposits, as proposed for Cerro Galan and some other large calderas (Sparks et al., 1985; Gudmundsson, 2008; Gregg et al., 2012; Cashman and Giordano, 2014). However, the cited geologic evidence for these partly model-based interpretations (absence of proximal fallout deposits beneath the ignimbrite, estimated large intracaldera volumes relative to the associated outflow tuff, and high viscosity of crystal-rich dacite magma) seems open to alternatives. Thick intracaldera accumulations of compositionally uniform ignimbrites document that the associated caldera collapsed at some stage during an eruption but provide no direct stratigraphic evidence for inception time; at least some large ignimbrites of crystal-rich dacite generated voluminous distal fallout deposits (Best et al., 2013a, 2013b); and some were highly mobile, comparable to crystal-poor rhyolites. For example, distal Fish Canyon Tuff, the archetypal “monotonous-intermediate” ignimbrite (Hildreth, 1981), is preserved along the Arkansas River Valley, 125 km northeast of the rim of La Garita caldera (McIntosh and Chapin, 2004, p. 228), and a similar distance south of the caldera at Las Tablas, New Mexico (Zimmerer and McIntosh, 2011). For the Bonanza ignimbrite, compositional variations demonstrate that caldera collapse began only after voluminous eruption of the eastern outflow ignimbrite. Later-erupted intracaldera tuff ponded to multi-kilometer thickness; intracaldera and outflow ignimbrite volumes appear roughly subequal, but proximal precursor fallout is absent.

Bonanza caldera was localized within the large edifice of the composite Rawley volcanic complex (ca. 33.7–33.3 Ma), which probably recorded early incremental growth and petrologic evolution of the upper-crustal magma body that culminated in an ignimbrite eruption. Similar precursor sequences are well documented elsewhere in the San Juan region and other Cordilleran ignimbrite centers (Lipman, 1984, 2007). Although even the most pristine rocks from the Bonanza system display effects of weak alteration, especially for mobile elements such as the alkalis, the analytical data (Supplemental Table 4) are sufficiently abundant to define chemical trends and contrasts with other parts of the SRMVF. Detailed petrologic study of the Bonanza eruptive sequence is currently in progress (e.g., Memeti and Lipman, 2014), and only brief aspects of magmatic evolution are noted here.

Rocks of the Bonanza cycle (Table 4) are more geochemically diverse and modestly more alkalic than younger volcanic assemblages associated with large ignimbrite eruptions farther south and west in the San Juan region (Fig. 22). In comparison to other San Juan ignimbrites, the Bonanza Tuff is characterized at any equivalent silica content by elevated incompatible elements such as total alkalis, Zr, Rb, La, and Ce (Fig. 22). The least altered samples of Bonanza Tuff, especially outflow dacitic phases, are also higher in alkalis and incompatible trace elements than otherwise compositionally similar precaldera and postsubsidence lavas and intrusions of the Bonanza cycle.

The Bonanza Tuff contains the broadest bulk-compositional (60%–76% SiO₂) range of any ignimbrite erupted from SRMVF calderas, and its crystal cargo is distinctive. Rhyolitic Bonanza Tuff is more silicic (to 76% SiO₂) than the other Oligocene ignimbrites, and its phenocrysts are nearly unzoned and in apparent equilibrium with a rhyolitic magmatic liquid. Crystal-poor rhyolite (2%–5% phenocrysts) contains sodic low-Ba (Cn_<0.1) sanidine of restricted compositional range (Or_51–54), along with accessory titanite and hornblende in some samples. In contrast, feldspars in dacitic Bonanza Tuff display complex disequilibrium zoning and resorption features (Figs. 23 and 24). Compositionally variable sanidine with much higher Ba contents (Or_55–70; Cn_<0.1–>6) is present both in bulk samples and in nonfragmented pumice fiamme of low-silica dacite (62%–64% SiO₂), a more mafic host composition than known for this mineral elsewhere in the region. The diversity of sanidine compositions and textures in dacitic Bonanza Tuff are similar but more extreme, in comparison to other dacitic ignimbrites and lavas of the SRMVF, indicating that magma mixing, pressure changes, and variations in volatile components occurred commonly during prolonged processes of magma assembly (Lipman et al., 1978; Bachmann et al., 2002, 2014).

Bulk dacite samples, although dominated by pumices of low-silica dacite, commonly also contain sparse dark scoria of silicic andesite (60%–62% SiO₂) that lack sanidine, and the lowest welded zone in the Findley Ridge section (Fig. 14) consists entirely of transitional andesite-dacite that lacks sanidine. Dacite zones also commonly contain minor proportions of light-gray pumices with only sparse crystals and more silicic appearance (but too small to sample and analyze). The diverse sanidine compositions, both in bulk-ignimbrite samples and individual pumice lenses (Fig. 24) also document both pre-eruption magma mingling and further mechanical mixing of rhyolite and dacite during eruption and emplacement processes. At least three discrete magma compositions are resolvable from pumice and bulk-sample compositions; evaluation of potential additional compositions transitional from dacite to silicic rhyolite is impeded by the small size and dense welding of rhyolitic pumices. Multiple pumice compositions have also been described from other SRMVF ignimbrites such as Carpenter Ridge, Sunshine Peak, and Grizzly Peak Tuffs (Lipman, 1975; Fridrich and Mahood, 1987; Hon and Lipman, 1989; Bachmann et al., 2014).

Such mixed pumice populations and bulk compositional variations have been diversely interpreted to record (1) variable drawdown depth during fluctuating eruption dynamics from a single compositionally zoned or layered magma body (Smith, 1979; Blake, 1981; Bacon and Druitt, 1988; Hildreth and Wilson, 2007), or alternatively (2) tapping multiple isolated magma pockets in close lateral or vertical proximity (Maughan et al., 2002; Eichelberger et al., 2000; Ellis and Wolff, 2012; Cashman and Giordano, 2014). For Bonanza and other SRMVF ignimbrites, the overall progression from early rhyolite to large volumes of later-erupted dacite, along with structural evidence for coherent near-equant subsidence and resurgence of a relatively intact intracaldera block (floor and ignimbrite fill), are here interpreted as primarily recording vertical compositional variations within a single relatively well-interconnected shallow lens of eruptible magma, rather than tapping multiple discrete bodies of more homogeneous magma (Fig. 15). The alternating rhyolite and dacite zones in the Bonanza Tuff are inferred to record fluctuating drawdown depths that tapped vertically stacked compositional zones in the source magma body, as the intensity of eruptive discharge pulsated or opening of new fractures during subsidence tapped different melt levels. The compositional and textural diversity of this ignimbrite suggests eruption from a crystal-poor silicic cap and underlying more crystal-rich dacite, only a few kilometers thick, that accumulated above more mafic and vertically extensive mushy magma, similar to that inferred for other ignimbrites (Smith, 1979; Hildreth, 1981; Bachmann and Bergantz, 2004).

The contrasting compositions between eastern and western areas of preserved outflow Bonanza Tuff (Fig. 11) could suggest an alternative mechanism, involving concurrent eruption of large discrete magma bodies that differed in composition and were laterally isolated (Eichelberger et al., 2000), as appears to have been common in the extensional environment of the Taupo zone (e.g., Gravley et al., 2007; Allan et al., 2012; Bégué et al., 2014). Similar interpretive issues have been much discussed for the 0.76-Ma Bishop Tuff, from Long Valley caldera (Hildreth and Wilson, 2007; Gualda and Ghiorso, 2013, and references therein). Such a geometry of multiple dispersed magma bodies seems less appropriate at Bonanza; however, Bonanza is a single large caldera without evidence for satellitic subsidence or vent sites, generation of the many compositional alternations within the ignimbrite would have required prolonged stable access to separate discrete magma bodies throughout the course of a sustained large-volume eruption, and the full spectrum of mafic to silicic magma continued to be available after the ignimbrite eruption on both flanks of the resurgent dome (lavas on the west and resurgent intrusions on the east). Also lacking at the time of the Bonanza eruption was any strong extensional environment such as characterizes Taupo.

The voluminous but short-lived postignimbrite magmatism at Bonanza, including caldera-filling lavas and resurgent intrusions, also was characterized by compositional and textural diversity (Table 4 and Fig. 22). Caldera-filling flows range from andesite (56% SiO₂) to silicic rhyolite (76% SiO₂), emplaced within ∼10⁵ yr after the ignimbrite eruption. Resurgent intrusions that were assembled and solidified in about the same time span similarly range from mafic granodiorite and andesite (54%–62% SiO₂) to aplitic granite (76% SiO₂). These later magmas of the Bonanza cycle, while also more alkalic than rocks of the central San Juan caldera cycles, are somewhat lower in incompatible elements than samples of Bonanza Tuff at similar silica contents (Fig. 22).

Preliminary radiogenic-isotope data for Bonanza rocks (Memeti and Lipman, 2014) indicate relatively limited ranges in Sr, Nd, and especially low Pb values, compared to the multiple-cycle ignimbrite eruptions from Platoro and central San Juan calderas farther south and west, consistent with the relatively brief duration of the Bonanza cycle as constrained by ⁴⁰Ar/³⁹Ar ages (Table 3). The Bonanza lead values are relatively low (²⁰⁶Pb/²⁰⁴Pb = 17.4–17.9), and their range is closer to those for the early-intermediate lavas that preceded caldera cycles than to ignimbrites elsewhere in the San Juan region. Such low lead values have long been interpreted as reflecting incorporation of lower crust depleted in U and Th during Proterozoic craton assembly (Lipman et al., 1978; Johnson, 1991; Riciputi et al., 1995).

The great compositional diversity and recurrence of andesite as magmatic products during the temporally brief caldera cycle at Bonanza document the capacity of large upper-crustal magmatic systems to evolve rapidly, especially when recurrently recharged by arrival of new batches of mafic magma. The dominant andesitic to rhyolitic magmas are interpreted to have been 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; Hildreth and Moorbath, 1988; Johnson, 1991; DePaolo et al., 1992; Riciputi et al., 1995; Farmer et al., 2008). The absence of erupted basalt, prior to onset of regional extension, shows that primitive mantle melts were unable to penetrate a developing upper-crustal batholith at Bonanza, probably because warm near-solidus batholithic rocks were a barrier to rise of hot mafic magma without attendant cooling, crystallization, and stagnation.

Bonanza caldera displays diverse structural and compositional features that provide special insights concerning ignimbrite eruptive processes. Bonanza, source of a compositionally complex regional ignimbrite sheet erupted at 33.12 ± 0.03 Ma, is a subequant structure ∼20 km diameter that subsided >3 km during eruption of ∼1000 km³ of ignimbrite. In contrast to the multicyclic caldera loci for many large-volume ignimbrite flareups elsewhere, Bonanza is an areally isolated collapse structure that formed in response to a single ignimbrite eruption. A large body of high-precision ⁴⁰Ar/³⁹Ar ages shows that the entire Bonanza caldera cycle, including ignimbrite, postcollapse lavas, most resurgence, and associated intrusions, occurred within the relatively brief interval from 33.12 ± 0.03 to 33.03 ± 0.04 Ma. Later erosion has exhumed some near-original caldera morphology that is expressed by present-day landforms (western topographic rim, resurgent core, and ring-fault valley), while tilting and deep erosion provide exceptional three-dimensional exposures of fill, floor, and resurgent structures. Caldera-fill ignimbrite has been largely stripped from the southern and eastern flank of the dome, exposing large areas of caldera floor as a coherently domed plate broken by only small-displacement faults, and bounded by ring faults with locations that are geometrically closely constrained even though largely concealed beneath valleys.

The Bonanza Tuff displays extreme compositional gradients (silicic andesite to rhyolite; 60%–76% SiO₂), multiple alternations of rhyolite and dacite zones rather than simple upward gradation from silicic to mafic, and compositional contrasts among outflow sectors (mainly crystal-poor rhyolite to east and crystal-rich dacite to west). Varied thickness and lateral extent among the multiple interfingering zones of rhyolite and dacite within the intracaldera ignimbrite reflect multiple factors, including surface irregularities on the pre-eruption lava assemblage, depositional slopes generated by caldera-wall landslide deposits, and weakly asymmetric caldera subsidence. The relatively modest thicknesses and limited areal extent of the lower-rhyolite zone within the caldera provide rigorous constraints on timing of initial collapse; it must have begun relatively late during this phase of the Bonanza ignimbrite eruption. If subsidence had accompanied inception of the eruption, or even triggered initial magma expulsion as inferred in some ignimbrite calderas, a greater thickness and volume of the early rhyolite should have ponded within Bonanza caldera.

For Bonanza and other SRMVF ignimbrites, the overall progressions from early rhyolite to later-erupted dacite, along with structural evidence for coherent near-equant subsidence and resurgence of a relatively intact intracaldera block (floor and ignimbrite fill), are interpreted as primarily recording vertical compositional variations within a single well-connected shallow lens of eruptible magma, rather than tapping multiple isolated bodies of more homogeneous magma. The alternating rhyolite and dacite zones in the Bonanza Tuff are inferred to record variable drawdown that tapped a vertical sequence of compositional zones in the source magma body as the intensity of eruptive discharge fluctuated.

Particularly revealing at Bonanza caldera are the widespread exposures of caldera-floor volcanic rocks and basal deposits associated with the ignimbrite eruption and caldera collapse. The structurally coherent floor at Bonanza contrasts with fault-disrupted floors at other well-exposed multicyclic calderas where successive ignimbrite eruptions caused recurrent subsidence. Notable at Bonanza are zones of floor-rock shattering and brecciation within ∼100 m inboard of ring faults, perhaps due to compression and crushing of the subsiding floor in proximity to steep inward-dipping faults adjacent to caldera-collapse ring faults. In-place caldera floor grades into caldera-fill megabreccia with boundaries that can be located only approximately. Dike-like ignimbrite crack fills of strongly welded tuff, which penetrate upper levels of the caldera floor and have steeply dipping fiamme, are interpreted to have originated mainly as surficially generated fills between blocks of early caldera-collapse megabreccia into cracks that opened dilatantly during caldera formation, rather than originating as ignimbrite vents.

The absence of tephra-fall deposits beneath the Bonanza Tuff and other SRMVF ignimbrites is interpreted to indicate early generation of pyroclastic flows, rather than evidence for lack of a Plinian eruptive column. A discharge rate that was high from inception of the eruption could have produced rapid ignimbrite flowage at the onset of the eruption, such that any proximal Plinian pumice and ash could fall directly into the moving pyroclastic flow, no basal fall layer would be deposited proximally, and tephra-fallout deposits would be preserved only distally beyond the ignimbrite sheet. Although the absence of a Plinian deposit beneath some ignimbrites elsewhere has been interpreted to indicate that abrupt rapid foundering of the magma-body roof initiated the eruption, initial caldera collapse began at Bonanza only after several hundred kilometers of rhyolitic tuff had erupted, as indicated by the small thickness and modest volume of the lower rhyolite within the caldera.

After completion of the ignimbrite eruption and accumulation of andesitic to rhyolitic lavas in the caldera basin, the floor and fill of Bonanza caldera were arched into an notably high and steep-sided dome by continued magma rise, forming plutons of granite and granodiorite that are exposed at roof levels. Resurgence uplifted the pre-ignimbrite caldera floor to levels as high as the elevation of the same lava units on the western caldera rim, leading to widespread present-day erosional exposure of caldera-floor rocks in the caldera interior. Estimates of intracaldera ignimbrite volume at little-eroded young calderas, based on morphology of resurgent uplifts without information on caldera-floor geometry, might yield misleadingly large estimated ratios of intracaldera versus outflow ignimbrite accumulations. The insights provided by Bonanza caldera and its ignimbrite should be broadly applicable to processes of silicic Cordilleran magmatism elsewhere, with respect to ignimbrite eruptive processes, style and timing of caldera subsidence, complications with characterizing structural versus topographic margins of calderas, contrasts between intra- versus extracaldera ignimbrite, scales of caldera resurgence, and the limitations in assessing magma volumes associated with large caldera-forming eruptions. Bonanza provides a rare site where both caldera margins and caldera floor are exhumed and exposed, providing valuable perspectives for understanding similar young calderas in some of the world’s most active and dangerous silicic provinces.

We especially thank Andrea Sbisà, who provided outstanding mapping assistance during long field days in the summers of 2009–2011. Lisa Peters and numerous students at the New Mexico Geochronology Research Laboratory helped with mineral separates and data collection. Detailed perceptive manuscript reviews were provided by U.S. Geological Survey (USGS) reviewer David John, Geosphere reviewers Mike Branney, John Dilles, and Darrel Gravley, and editors Jan Lindsay and Shan de Silva. We also thank many friends in the northeastern San Juan region who provided diverse hospitality, logistical support, help with back-country and property access, and other assistance. These include John and Patty Judson of Quarter-Circle Circle Ranch in Cochetopa Park; Forest, Billy, Curt, and Lee Ann Cadwell of Cathedral Creek Ranch; and especially for the recent years of Bonanza studies, Pip and Aaron Conrad of the Rafiki Ranch near Villa Grove, who have created a virtual “Rafiki Ranch Geologic Research Station” and hosted many visiting earth scientists.

Geologic mapping of western fringes of the Bonanza area, initiated to resolve interpretive problems within the adjacent Cochetopa–North Pass calderas (Lipman, 2012; Lipman et al., 2013), was gradually expanded to cover Bonanza and adjacent areas to the northeast (Fig. 2) as multiple stratigraphic, structural, geochronologic, and volcanologic complexities emerged. Detailed geologic mapping of the Bonanza-Marshall caldera area (field compilation on 1:24,000 topographic quadrangle base maps; intended for publication as a USGS SIM-series map, 1:50,000 scale), is a primary basis for documenting many of the interpretations developed here.

Critical to many interpretations presented here are the high-precision ⁴⁰Ar/³⁹Ar age determinations for volcanic rocks in the Bonanza area (130 localities; 138 mineral and groundmass ages: Table 3; Supplemental Tables 1–3), including 28 determinations for regional units that were published previously (McIntosh and Chapin, 2004; Lipman and McIntosh, 2008). All ages were determined at the New Mexico Geochronology Research Laboratory by methods similar to those described in McIntosh and Chapin (2004); additional details on sample preparation and irradiation, instrumentation, analytical parameters, age calculations, and probability plots are listed in Supplemental Tables 2 and 3. Especially high precision ages were obtained for samples analyzed with the multi-collector Argus VI mass spectrometer, starting in 2012. A single-detector MAP 215-50 was used to determine isotopic ratios prior to 2012. All new and previously published ages are calibrated to Fish Canyon Tuff (sanidine) at 28.02 Ma, for consistency with prior reports. All results are presented with 2σ analytical uncertainty (95% confidence interval). Interpreted preferred ages (Table 3 and Fig. 4) are influenced by stratigraphic relations, supplemented by additional age determinations from beyond the map area; interpretive problems with ages for some units are discussed by Lipman and McIntosh (2008).

Especially useful for interpretations are single-crystal, laser-fusion ages of sanidine, obtained wherever this phase was present. Xenocrysts, sanidines with anomalously low radiogenic yields, and plagioclase crystals were excluded from weighted-mean age calculations. Pooled sanidine ages for multiple sites from a single unit are listed as weighted means, with uncertainties as the standard error of the mean (S_e).

Additional determinations are multi-crystal, step-heating ages for biotite and hornblende phenocrysts, K-feldspar from some granitoid rocks, and groundmass concentrates from lavas lacking datable phenocrysts. For the step-heated samples, plateau ages are calculated when the age spectrum contains three or more contiguous steps that comprise >50% of the ³⁹Ar released, and the steps overlap at 2-sigma (Fleck et al., 1977). Ages were calculated using the inverse isochron for samples that contained excess ⁴⁰Ar (⁴⁰Ar/³⁶Ar > 295.5; Nier et al., 1950). Many of the isochron ages agree, within 2σ analytical uncertainty, with plateau or integrated ages for the same sample, but some are younger. Excess ⁴⁰Ar is especially common in the intrusive samples from Bonanza, probably reflecting a volatile component released from the upper-crustal Proterozoic wall rocks when heated by the growing subvolcanic magma bodies. The isothermal duplicate step-heating technique was used to remove excess ⁴⁰Ar from fluid inclusions in K-feldspar (Harrison et al., 1994). High-temperature steps also yielded anomalously old ages, likely related to excess ⁴⁰Ar in the largest domains (Foster et al., 1990). Plutonic K-feldspar from Questa (Zimmerer and McIntosh, 2012a) and Mount Aetna calderas (Zimmerer and McIntosh, 2012b), similarly exposed caldera systems in the Southern Rocky Mountain volcanic field (SRMVF), also contains significant excess ⁴⁰Ar.

Some biotite, hornblende, and groundmass concentrate step-heating analyses yielded discordant spectra interpreted to be a combination of minor alteration of interstitial glass and ³⁹Ar recoil related to neutron radiation. Results are reported as integrated (or total-gas) ages if the spectrum fails to meet the criteria for a plateau age and the inverse isochron does not display a coherent array of data. Particularly problematic have been anomalously young groundmass ages from some fine-grained intermediate-composition lavas that lack datable phenocryst phases. Several discordant spectra contain significantly young low-temperature steps that increase in age during the first ∼10%–20% of the ³⁹Ar released. Despite petrographic screening to avoid glassy matrix material, the young ages probably reflect ⁴⁰Ar loss from low retentive sites, such as poorly crystallized or otherwise altered interstitial sites where potassium was concentrated. Inclusion of these young, low-temperature steps commonly yields integrated ages with large uncertainties and, in some cases, ages that conflict with the established stratigraphy.

Chemical and petrographic data for the study area include new major-oxide and trace-element analyses for about 280 samples, determined by X-ray fluorescence methods (Johnson et al., 1999) at the Geoanalytical Lab, Washington State University (Table 4; Supplemental Table 4), supplemented by data for regional units (Lipman, 2006, 2012). Sample selection for analytical work was complicated within Bonanza caldera, due to widespread propylitic metamorphism of deeply buried volcanic rocks, along with local acid-sulfate hydrothermal alteration and supergene leaching of disseminated pyrite. Accordingly, wherever possible, units were sampled from outflow areas beyond the caldera. Because the ignimbrites are typically densely welded, most analyses are for bulk samples; only a few localities of dacitic Bonanza Tuff contain pumice lenses sufficiently large to sample and analyze separately (noted on Table 4). During preparation for analyses, bulk ignimbrite samples were coarse crushed, and any discernable lithic fragments were discarded. All major-oxide analyses were recalculated volatile free, to sum to originally reported analytical totals. Despite potential alteration issues, most samples yielded plausible magmatic values, even for especially mobile elements such as the alkalis (e.g., Fig. 22). Effects of alteration are evident, however, in the modestly greater data scatter on data plots than typical of young volcanic suites. A few obviously anomalous samples with suspect compositions (marked in red, on Supplemental Table 4) were omitted from plots of chemical data.

Mineral compositions determined by routine electron-microprobe analysis on polished thin sections (Lipman, unpubl. data, 2006-2011; methods similar to those of Lipman and Weston, 2001), have proved useful discriminants to test correlations among some ignimbrite sheets, as well as providing information on processes of magmatic evolution.