Precursors to a continental-arc ignimbrite flare-up: Early central volcanoes of the San Juan Mountains, Colorado, USA

Continental-arc volcanism generates the largest and potentially most hazardous explosive eruptions known on Earth; some individual ignimbrite sheets have volumes >10³ km³. Such ignimbrites have been much studied during the past 50+ yr along the Cordilleran margin of the Americas. In addition to the ignimbrite flare-up discussed in this paper, well-documented examples include the large eruptive centers in the Basin and Range of Utah and Nevada (Henry and John, 2013; Best et al., 2016), the mid-Cenozoic Mogollon-Datil volcanic field in western New Mexico (Elston, 1984; Chapin et al., 2004), the vast Sierra Madre Occidentale in northern Mexico (McDowell and Clabaugh, 1979; Ferrari et al., 2007), and the younger Altiplano-Puna volcanic complex of northern Chile (de Silva et al., 2006). The widespread areal extent of such ignimbrites provides datable stratigraphic markers that have been critical to constrain eruptive histories of volcanic fields; their silicic compositions with phenocrysts of sanidine and/or biotite permit high-precision ⁴⁰Ar/³⁹Ar dating; their large volumes provide petrologically informative quenched samples of batholithic-scale upper-crustal magma bodies; and structures formed during associated caldera collapse have localized ore deposits and geothermal systems.

Much less studied have been the varied eruptive precursors to the continental-arc ignimbrite flare-ups, in part because such deposits have been widely eroded, concealed beneath outflow ignimbrite sheets, or caved away within calderas. Such early eruptions tend to produce many small-volume lavas and associated volcaniclastic deposits of limited areal extent that have been challenging to map in detail, to develop reliable laterally extensive stratigraphic control, or even to determine the locations and geometry of individual eruptive centers. Their dominant intermediate compositions (andesite, mafic dacite) carry typical crystal cargoes (plagioclase, pyroxene ± hornblende) that can be difficult to date reliably. Deep burial of some deposits has also resulted in low-grade thermal alteration that obscures magmatic compositions. Nevertheless, such volcanoes can provide critical records for events preceding large ignimbrite eruptions, including petrogenetic processes during early assembly of subvolcanic magma bodies, indications of local and regional tectonic stress during magmatism, and records of the tempo and compositional progression of eruptions leading to the culminating pyroclastic activity. Volcanism precursor to ignimbrite eruptions has increasingly been recognized as providing records of translithospheric magmatic processes during prolonged assembly of large source magma reservoirs (Lipman et al., 1978, 2022c; Townsend et al., 2019; Bouvet de Maisonneuve et al., 2021; Hildreth et al., 2023).

Here, we summarize and compare new and published stratigraphic, geochronologic, and petrologic data for the pre-ignimbrite volcanoes within the San Juan locus of the mid-Cenozoic Southern Rocky Mountain volcanic field (SRMVF), Colorado (Fig. 1; Table 1). The early volcanism is exceptionally widespread, voluminous, and well preserved in the San Juan Mountains, constituting the dominant volume of eruptive deposits at this major site of continental-arc ignimbrite volcanism (Lipman et al., 1970, 1978; Steven and Lipman, 1976; Lipman, 2007). Although a few reports have provided petrogenetic interpretations for individual early centers and made comparisons with other available data (Colucci et al., 1991; Parker et al., 2005), this overview represents the first regional summary of early San Juan volcanism since pioneering field and petrologic studies conducted before ignimbrites were distinguished from lavas (Cross and Larsen, 1935; Larsen and Cross, 1956).

We review field stratigraphic and map relations, age control on growth histories, and compositional evolution for individual early eruptive centers, with special focus on recent results from eight eastern San Juan volcanoes. Individual constructional edifices are defined by proximal accumulations of lavas and breccias, central intrusions, and outward-trending dikes. These are especially well preserved and exposed in the eastern San Juan region because many lie outside the margins of later ignimbrite calderas, are at relatively low elevations, and, as a result, are less obscured by vegetative cover. These volcanoes erupted broad ranges of magmatic compositions, and some contain phenocrysts, including sanidine, that permit high-precision age determinations.

The mid-Cenozoic SRMVF, site of 30 large ignimbrites (mainly 37–27 Ma) and related calderas and subvolcanic intrusions (Fig. 1), provides an exceptional laboratory for studying processes of continental-arc magmatism (Lipman, 2007; Lipman and Bachmann, 2015; Best et al., 2016). In places, virtually pristine volcanic morphology has been exhumed by recent erosion; elsewhere, rugged topography and structural tilting expose multikilometer volcanic sections, down into upper levels of subvolcanic intrusions.

Dominantly intermediate-composition lavas and associated breccias (andesite, dacite) were voluminous precursors to the ignimbrite eruptions, and eruption of similar lavas continued concurrently with the major ignimbrites, commonly filling caldera depressions (Lipman, 2007, and references therein). Although the early volcanic centers were widely scattered, several clusters became sites for later ignimbrite eruptions and attendant calderas. The clustered centers have long been interpreted as recording early magmatic focusing and initial growth of sizable upper-crustal magma bodies capable of erupting large ignimbrites (Lipman et al., 1978; Lipman, 1984; Lipman and Bachmann, 2015). While substantial uncertainties remain, major ignimbrite eruptions and related calderas appear to have occurred substantially after peak growth of spatially associated precursor volcanoes—as much as several million years later (Lipman et al., 2022c).

Volcanic centers of the SRMVF, both the early intermediate-composition lavas and later ignimbrites, tended to migrate from north to south, from the present-day Sawatch Range southward into a broader locus in the San Juan Mountains and into the Questa-Latir area of northern New Mexico (Fig. 1; Lipman, 2007). The general southward migration is parallel to that long documented for eruptions in the Basin and Range region, probably related to disruption of the subducted Farallon plate (Stewart et al., 1977; Lipman, 1980; Henry and John, 2013; Best et al., 2016). Such arc volcanism is widely interpreted as initiated by basaltic magma from the mantle, but mid-Cenozoic basaltic lavas were largely absent prior to and during the ignimbrite eruptions in the SRMVF, despite persistent search for primitive compositions (Lake and Farmer, 2015). Their absence is inferred due to lower-crustal assimilation and fractionation (Lipman et al., 1978; Riciputi et al., 1995; Farmer et al., 2008).

The earliest regional ignimbrites of the SRMVF erupted from caldera sources aligned along the present-day Sawatch Range (Fig. 1). Proximal volcanic rocks have been largely eroded from northern parts of the Sawatch trend, and distal outflow ignimbrites and associated lavas are preserved mainly to the east, in the Thirtynine Mile volcanic area. Early lavas and breccias, overlain by ignimbrites, are increasingly thick and well preserved southward into the San Juan Mountains, which contain the largest erosional remnant of the composite SRMVF (Cross and Larsen, 1935; Steven et al., 1974). At the San Juan locus, the early volcanoes and their clastic aprons (~25,000 km³) constitute almost two thirds of total volcanic volume (Lipman et al., 1970; Lipman, 2007).

Exposure levels of volcanic rocks along the east flanks of the Sawatch Range and San Juan Mountains are influenced by contrasts in synvolcanic erosion and postvolcanic tilting associated with extension along the Rio Grande rift. Volcanic rocks along the Sawatch trend dip gently westward as far south as the town of Saguache, while strata farther south are tilted eastward all the way into northern New Mexico (Fig. 1). These regional tilts, recorded especially clearly by the ignimbrite sheets, result in contrasting exposure levels for eastern and western flanks of the early volcanoes discussed here.

The sequences of early intermediate-composition lavas that preceded ignimbrite eruptions have been assigned multiple local stratigraphic names in diverse areas of the SRMVF. For the San Juan locus, these have generally been included in the regional Conejos Formation, subdivided by volcanic edifice and compositional grouping where studied in sufficient detail (Table 1). Locations of some early volcanoes were crudely defined during twentieth-century regional mapping (Cross and Larsen, 1935; Steven et al., 1974) by central intrusions and outward-projecting dikes. Abundance of proximal lavas and breccias relative to distal volcaniclastic rocks provides further evidence of volcano geometry.

Volcanic rocks of the San Juan locus were largely deposited along the eastern margin of the Colorado Plateau where mid-Cenozoic topography was subdued, and the outflow ignimbrite sheets, even though widely only a few tens of meters thick, maintain lateral continuity and thickness for long distances, some still preserved more than 100 km from source calderas. In contrast, pre-ignimbrite central volcanoes of the Conejos Formation consist of many individual lava flows, necessarily involving complex lateral stratigraphic interfingering and local thickness variations.

Conejos lavas and breccias commonly are a kilometer or more thick in the San Juan locus, as well diagrammed in regional cross sections (Larsen and Cross, 1956, their plate 2). Complete stratigraphic sections through the thick Conejos Formation are locally well exposed, especially along the steep southwestern erosional flanks of San Juan Mountains, where cliff-forming lava and breccia sequences rise as much as 1300 m from the Mesozoic-floored San Juan Basin to the crest of the Continental Divide (Table 2). Farther east, comparable and greater subsurface thicknesses, to 2500 m, have been documented by petroleum-exploration drill holes (Curtis, 1988; Gries, 1985; Brister and Gries, 1994). Lithologies near the base of the volcanic deposits that are suitable for reliable age determinations are frustratingly scarce, but scattered locations in the eastern and central San Juan region have mostly yielded broadly similar ages at ca. 35 Ma for inception of the pre-ignimbrite volcanism (Table 3).

In the northeastern San Juan Mountains along the southern end of the Sawatch trend (Fig. 1), the thickness of the early lavas is more variable due to rugged prevolcanic paleotopography on early Cenozoic (Laramide) uplifts. In places, ignimbrite sheets were even deposited directly on Proterozoic crystalline rocks (Lipman, 2020). Stratigraphic relations between lavas and ignimbrites also are more complex in the northeastern San Juan region, where early ignimbrites (34–32 Ma) erupted from sources along the Sawatch trend are equivalent in age to lava accumulations of the Conejos Formation that underlie all ignimbrites of the San Juan locus to the south and west (Fig. 2). Despite excellent exposures in many areas with topographic relief locally up to 1500 m, interpretation of the eruptive history and overall geometry of individual eruptive centers is further complicated by synvolcanic erosion of upper parts of volcanic edifices and inaccessible lava sequences in the subsurface.

Younger andesitic and dacitic lavas are also interstratified with the San Juan ignimbrites and ponded in associated calderas. These lavas are compositionally little different from the earlier-erupted Conejos assemblage. The volcanic history of the San Juan locus, and more broadly the entire SRMVF, can thus be summarized as a multi-million-year sequence of many intermediate-composition lava eruptions, individually of small volume, punctuated at late stages by ignimbrite supereruptions and associated calderas.

This article especially focuses on early central volcanoes along the northeast flank of the San Juan Mountains, all included in the Conejos Formation. Several of these were recognized early in the twentieth century (Cross and Larsen, 1935; Larsen and Cross, 1956), but little further field and analytical data had been obtained before the early 2000s, other than the well-exposed Summer Coon volcano characterized by spectacular radial dikes (Lipman, 1968, 1976; Mertzman, 1971). Here, we report new stratigraphic, petrologic, and geochronologic data for two prominent pre-ignimbrite volcanic centers (Biedell-Lime, Baughman) that previously have been little studied, and we summarize published results for other early eruptive centers of the San Juan locus and the southern end of the Sawatch trend (Fig. 1; Table 1). Map and analytical data for many of the early volcanoes were obtained during studies focused on ignimbrite eruptions and attendant caldera formation (Lipman, 2000; Lipman and McIntosh, 2008; Lipman et al., 2015), but prior interpretation has been limited. In addition to early volcanoes east and northeast of the large 28.2 Ma La Garita caldera, we summarize age and compositional results for pre-ignimbrite centers at the multicyclic Platoro caldera complex (30.2–28.8 Ma) to the south, the 33.3 Ma Bonanza caldera to the north, and limited data for precursors to ignimbrite eruptions from western San Juan calderas.

We mapped the previously little-studied Biedell-Lime and Baughman volcanoes at a scale of 1:24,000, mainly during 2006 and 2017–2019, distinguishing petrologically distinctive lava packages, dikes, and other intrusions, and demarking areas of hydrothermal alteration. The only published maps for these eruptive centers, based on reconnaissance at 1:250,000 scale (Cross and Larsen, 1935; Steven et al., 1974), are highly generalized; only a few volcanic units were distinguished, and some features were incorrectly located even at the small map scale. Published age and chemical analytical data were sparse (a single 1960s-era K-Ar date, three major-oxide analyses from the early twentieth century; Lipman et al., 1970; Larsen and Cross, 1956). Six additional chemical analyses from southern parts of the Biedell-Lime center (described as the Carneros Creek volcano) were presented by Parker et al. (2005).

Field mapping of these two volcanoes was supplemented by new and published analytical data for these and other early centers, including: ⁴⁰Ar/³⁹Ar (97 new, 109 published) and U-Pb zircon (9 new) age determinations (Table S1¹), petrographic studies, and 312 major- and trace-element chemical analyses (Table S2). In total, 114 volcanic and 62 intrusive samples have been dated. New geochronologic analytical data are summarized in Tables S3 and S4, and a brief history of San Juan age determinations, analytical techniques for new determinations, and interpretive issues for some results are discussed further in the Appendix. All new and previously published ⁴⁰Ar/³⁹Ar ages were calculated to a standard age of 28.201 Ma for the Fish Canyon Tuff (Kuiper et al., 2008). Throughout the article, reported ages were obtained by the ⁴⁰Ar/³⁹Ar method, except as indicated otherwise.

As discussed further in the Appendix, determination of accurate and precise eruption ages has been challenging for the early volcanoes because many of the intermediate-composition rocks lack reliably datable minerals. Sanidine, the most suitable mineral for high-precision ⁴⁰Ar/³⁹Ar analysis, has been found only in a few silicic Conejos units from the northeastern San Juan region. Some groundmass determinations yielded anomalously young dates that are inconsistent with stratigraphy, presumably due to varied ⁴⁰Ar loss from poorly crystallized material. Some biotite and hornblende ages from andesites and dacites are inconsistent with stratigraphic relations or where both were dated from the same sample or eruptive unit. Biotite and hornblende ages tend to be anomalously old, typically by a few hundred thousand years, but some by 1–2 m.y., for the few samples where sanidine could also be dated, and thus the ⁴⁰Ar/³⁹Ar dates of these phases are conservatively interpreted as maximum eruption ages. For such geochronologic issues, “geologic” complexities of the samples are likely more substantial than analytical uncertainties.

The U-Pb dates record time(s) of zircon crystallization, and interpretation of their age significance can be challenging, as has been much discussed. Volcanic U-Pb dates commonly are older than the eruption age determined by ⁴⁰Ar/³⁹Ar analyses, and interpretations of U-Pb intrusion dates are also complex, some recording protracted intervals of crystallization (even to the million-year range) between initial stages of magma-body assembly and final solidification (Miller et al., 2007; Lipman and Bachmann, 2015, and references therein).

A NNE-trending alignment (~10 km) of at least three intrusive loci and adjacent lavas along the eastern mountain front between Del Norte and Saguache (Fig. 3) is here designated the Biedell-Lime volcanic complex (BLVC), retaining the somewhat geographically inconsistent usage in prior brief summaries (Cross and Larsen, 1935; Larsen and Cross, 1956; Steven et al., 1974). Proximal lavas associated with one central intrusion were described as the Beidell (sic) quartz latite by Larsen and Cross (1956). We interpret the BLVC as including additional intrusive loci farther north and south, with associated dikes and diverse lavas ranging from basaltic andesite to rhyolite.

The three interfingering loci are characterized by distinctive phenocryst mineralogy, similar chemistry, and a unified late eruptive history. Early activity was mainly at the two peripheral loci then becoming more localized in a central area, followed by late eruptions of dacite along the flanks of the entire BLVC. Hornblende andesite is a widespread composition, in both lavas and dikes, in contrast to the largely anhydrous phenocrysts in andesites at other volcanoes to the south and north along the eastern mountain front. Several intrusions at the BLVC include small masses of monzonite that grade into fine-grained intrusive andesite and dacite, along with dikes of andesite, dacite, and rhyolite. The duration of activity, both eruptive and intrusive, was from ca. 35 to 33 Ma, the younger exposed rocks located centrally within the composite BLVC (Fig. 4). Overall geometry and growth history in this area are incompletely understood because much of the overall volcanic section is confined to the subsurface. A deep drill hole, the WECO/UPRC 21–8 Hellgate well on the southwest flank of the BLVC, penetrated 2540 m of Conejos–type rocks before bottoming in prevolcanic arkose of the Eocene Blanco Basin Formation (Table 2; Curtis, 1988; Brister and Gries, 1994).

Intrusive loci and associated volcanic units of the BLVC are described from southwest to northeast, corresponding roughly to increasing complexity and interpretive uncertainty. Especially in northern sectors, lavas of varied composition, perhaps from differing sources, overlap and interfinger in ways that make assignment to intrusive loci ambiguous.

A southern magmatic locus within the BLVC is defined by a core of intrusions and associated epithermal mineralization-alteration along low-relief slopes 1–2 km south of the Lime–Biedell Creek junction (Fig. 3). Small exposed areas of equigranular monzonite grade into aphyric intrusive andesite just south of an equant area of hydrothermally altered andesite ~1 km across. More intrusive and associated altered volcanic rock is likely concealed beneath widespread slope wash to the west. Dikes, mainly hornblende andesite, radiate southwest to south from this locus, some projecting as much as 6 km to exceptional distal exposures along Carnero Creek and beyond (Fig. 5A). Despite the radiating pattern, the dikes define a dominant overall NNE trend. Additional small dikes may not be exposed, especially the relatively thin andesites, but comparable concentrations of dikes are absent on the northeast side of the locus.

Host rocks for the dikes are gently dipping lavas and interfingering laharic and debris-avalanche deposits, composed mainly of relatively silicic hornblende andesite (60–61% SiO₂; 10–20% phenocrysts of plagioclase [Pl] ≥ hornblende [Hbl] ± clinopyroxene [Cpx]). These rocks dip gently outward from the intrusive locus, ~10° on the east and southeast flanks, but nearly horizontal on the west side, reflecting late eastward tilting on the passive west shoulder of the asymmetrical Rio Grande rift graben. Similar lavas of hornblende andesite are present as far south as La Garita Creek, where they are onlapped by the north flank of Summer Coon volcano (Lipman, 1976). Hornblende ⁴⁰Ar/³⁹Ar ages for both the southern andesitic lavas and crosscutting dike are relatively old, mainly in the range 35–34.5 Ma (Fig. 4).

Overlying the hornblende andesite on the southeast flank of the southern locus, there is pyroxene andesite (~56% SiO₂; 5–15% Pl > Cpx), mainly proximal breccia at least 200 m thick. These relatively mafic breccias with anhydrous phenocrysts are compositionally similar to the lower rocks that form the core of Summer Coon volcano to the south, but their distribution and dip show that they are flanks of the southern locus of the BLVC.

The southern andesite sequence, and also more northerly portions of the BLVC, is capped by erosional remnants of thick biotite-dacite flows (most 63–66% SiO₂; 15–30% Pl > biotite [Bt] > Cpx ± Hbl). In places, the southern andesitic shield was deeply eroded before dacite eruptions. A spectacular cross section through a late valley-filling dacite is exposed at the Hellgate narrows along lower Carnero Creek (Fig. 5B), and farther north, this same flow has developed columnar cordwood jointing along a near-vertical marginal contact (Fig. 5C). These dacites yielded biotite ages of 33.38 ± 0.03 and 33.23 ± 0.02 Ma (Table S1A), providing an upper bracket for apparent long-lived multistage volcanism at the southern locus.

A groundmass-concentrate age of 33.63 ± 0.20 Ma from the main intrusive andesite and a biotite age of 33.24 ± 0.02 Ma from a sizable dacite intrusion adjacent to the central altered area are also consistent with late magmatism of the southern locus at ca. 33.5 Ma (Fig. 4). In contrast, a U-Pb zircon date from the southern monzonite appears anomalously old, with a weighted mean of 36.3 ± 0.6 Ma, seemingly recording protracted pre-emplacement crystallization (antecrysts). An alternative, deemed less likely by us, could be an older intrusion that had been erosionally unroofed prior to emplacement of the compositionally similar intrusive andesite and adjacent dated lavas. Additionally, 11 of the 29 analyzed zircon crystals from the monzonite yielded Proterozoic ages, documenting survival of much older xenocrysts without dissolution in mid-Cenozoic magma. The xenocrystic zircons were not distinguishable in backscattered electron images from those with mid-Cenozoic ages.

Another magmatic locus, centered along Sanderson Gulch, is defined by a bull’s-eye configuration of variably oxidized, light-tan rhyolite lava ~3 km across, onlapped by flows and volcaniclastic deposits of hornblende andesite (Fig. 3). No base of the rhyolite mass is exposed, and lenses of green-gray vitrophyre probably are pressure-ramp features along upper levels of the original lava dome. Most of the rhyolite contains only a few percent plagioclase; a small southeastern area of more crystal-rich lava (70.7% SiO₂) along the Sanderson-Cottonwood divide contains ~15% plagioclase and biotite. No sharp contact was located with the more widespread crystal-poor body, but the crystal-rich lithology may mark a late-emplaced vent phase. Biotite from this phase yielded a ⁴⁰Ar/³⁹Ar date of 33.46 ± 0.05 Ma, unexpectedly young for the lowest exposed unit of the central locus.

A geographically and lithologically distinct small area of crystal-rich sanidine-bearing rhyolite lava (73.3% SiO₂; 20% Pl > sanidine [Sa], quartz [Qz] > Bt), along upper Lime Creek southwest of the more widely exposed central body of rhyolite, has a slightly older sanidine age of 33.60 ± 0.07 Ma. Only upper parts of the Lime Creek rhyolite body crop out at present erosion levels; its extent may be much larger in the subsurface. Located near the inferred interfingering boundary between lavas of the southern and central loci, it is inferred to represent an early phase of the central locus, along with the larger rhyolite mass to the northeast. Both rhyolites are younger than hornblendic lavas of the southern locus.

Hornblende andesites that overlie the central-locus rhyolite extend at least as far north as San Juan Creek (Fig. 3); to the south, these lavas merge with petrographically similar older flows of the southern locus. The andesite of the central locus must be younger than the underlying rhyolite along San Juan Creek, with its reliable sanidine age of 33.68 ± 0.11 Ma; andesite lava along Biedell Creek yielded a hornblende age of 33.26 ± 0.06 Ma. However, during field studies, no mappable contact was identified between the central andesite and the similar lithology of southern lavas that appear to be a million or more years older (34–35 Ma). Just as for the southern locus, lavas of crystal-rich dacite are present along flanks of the central locus at the edge of the San Luis Valley to the east and along the high western ridge crest that forms the topographic divide with the Carnero Creek drainage. Four biotite and sanidine dates from these lavas are in the range 33.2–32.9 Ma, similar to those for the upper dacites that flank the southern locus.

Diverse intrusions are associated with the central locus. The largest clearly intrusive body is a geometrically complex mass of finely porphyritic andesite-dacite (~62% SiO₂; 10%– 15% small Pl > Bt) that caps the ridge north of Sanderson Gulch (Fig. 3). Steep lower parts of this body truncate the lower rhyolite and mushroom upward into a lenticular thick sill or laccolith, roughly along the boundary between the lower rhyolite and overlying andesite. The Sanderson intrusion is massive and structureless, with an exposed thickness of as much as 200 m, its top eroded. Most of this body is fine grained, dense, and characterized by rectilinear joints, without internal structure. Such rectilinear jointing is typical of aphanitic phases of shallow intermediate-composition intrusions elsewhere in the San Juan region (Lipman et al., 2015, p. 1927). As the intrusive contacts steepen southward, the andesite-dacite grades into a NE-trending dike-like body of medium-grained quartz-monzonite (67% SiO₂) south of Sanderson Gulch. The quartz monzonite yielded a U-Pb zircon date of 33.0 ± 0.4 Ma, and the finer-grained body to the north has an indistinguishable ⁴⁰Ar/³⁹Ar biotite age of 33.14 ± 0.06 Ma. This body, shown as intrusive by Cross and Larsen (1935) and Larsen and Cross (1956), was mapped as lava by Steven et al. (1974).

The high ridge between Biedell Creek and Sanderson Gulch just to the southwest, locally known as “Crystal Hill” (Fig. 3), is capped by a different dacite body, even more silicic (68.6% SiO₂; 15% Pl > Bt, Sa). This body also has a gently dipping basal contact against underlying rhyolite lava, and discontinuous intervening andesitic volcaniclastic rocks, as beautifully exposed in the open-pit excavations of the Crystal Hill mine (Fig. 5D). In contrast to the massively structureless intrusion across Sanderson Gulch to the north, however, the Crystal Hill body displays contact-parallel flow layering at its base and steeply dipping, ramp-like flow layering in higher outcrops. Such flowage features, typical of viscous silicic lava flows, suggest to us that the dacite is the eroded remnant of a large lava dome. No flow breccia is exposed along the basal contact; either this typical lava facies was not developed or preserved in such a proximal environment, or, alternatively, the Crystal Hill body is partly endogenous. This bulbous dacite lava was mapped as intrusive by Steven et al. (1974).

A mass of monolithologic breccia, consisting entirely of dacite clasts that are petrologically indistinguishable from the ridge-capping flow-layered dacite, is exposed centrally in the open-pit mine excavation. This body was described by mine geologists as a breccia pipe within a stock (Pansze, 1987; Cappa and Wallace, 2007). The fragments are mostly slabby, commonly with dimensions of 3–30 cm. The breccia body is reported to have plan dimensions of 200 by 140 m, extending at least 300 m below the surface, as indicated by drill-hole data (Pansze, 1987). We suggest that this mineralized breccia may have originated as vent fill early during eruption of the dacite lava. An apparent feeder dike of petrographically similar dacite, which strikes northeast and merges with the overlying Crystal Hill lava dome, also is exposed low in the open pit, cutting through the breccia.

Two sanidine dacite ages are within uncertainties at 33.22 ± 0.06 Ma and 33.26 ± 0.13 Ma for a breccia clast and the overlying lava. The dike of silicified dacite that cuts upward through the breccia (Fig. 5D) yielded a weighted-mean U-Pb zircon date of 34.1 ± 0.4 Ma that is slightly older (Fig. 4; Table S3A), but the oldest zircons in the population may record antecrystic growth, thus potentially skewing the weighted mean age.

North and south of these two large bodies of the central locus, small andesite, dacite, and rhyolite dikes have a weak radial geometry but preferentially trend to the NNE. Some mapped dikes, for example, those east of Crystal Hill, are only exposed in construction excavations, and more dikes than have been mapped, especially relatively small andesites, are undoubtedly present.

The Crystal Hill mine, explored as an underground vein system in the late nineteenth century and then operated as an open pit in 1984–1986, is probably the most productive mine developed at any of the early intermediate-composition volcanoes in the San Juan region. Quartz crystals (including amethyst) were abundant in upper portions of the now-excavated breccia, giving the district its name. Gold and silver minerals were reportedly located mostly below the zone of quartz crystals and were associated with manganese oxide in the matrix of the breccia. Some 27,000 oz (760 kg) of gold and 40,000 oz (1100 kg) of silver were recovered from the modern Crystal Hill Mine (Cappa and Wallace, 2007).

The northern locus is the least defined of the three in the BLVC. It is characterized mainly by a large area of the crystal-rich dacite, which underlies the widespread andesites of the BLVC, and by a cluster of dacitic intrusions along the upper reaches of Little Cottonwood Creek (Fig. 3). Andesite lavas and breccias, including hornblende-bearing, coarsely plagioclase-phyric, and near-aphanitic types, lap out against the crystal-rich dacite, which forms rugged cliffs in Red Rock Canyon. The diverse andesite types were not distinguished in detail within the northern locus; compositions appear to interfinger rather than defining mappable stratigraphic sequences over any sizable area.

The dacite of Red Rock Canyon (Lipman, 2020), a distinctive reddish-brown to gray crystal-rich dacite (64–68% SiO₂; 20%–30% Pl >> Bt > Cpx), is the oldest volcanic rock exposed along the north flank of the BLVC. Thick accumulations of this dacite crop out over an equant area 6–8 km across, extending from Cottonwood Creek as far north as Tracy Creek (Fig. 3). Gray dacite that forms massive outcrops in lower San Juan Creek has similar phenocryst contents but is characterized by locally abundant small inclusions of incompletely mingled fine-grained andesite. In some outcrops, elongate andesite inclusions in flow-foliated matrix dacite appear misleadingly similar to lithic-bearing welded tuff. These thick lavas, which were previously inferred to constitute an early south flank of Tracy volcano (Lipman, 2020), now seem likely to define flanks of a separate northern magmatic locus of the BLVC. Northern areas of the dacite dip northerly toward Tracy Creek, while eastern areas along the mountain front at San Juan Creek dip to the east and southeast, and southwestern areas along Cottonwood Creek dip gently in that direction.

Two samples from dacitic lava south of Tracy Creek (65.0% SiO₂; ~25% Pl >> Bt > Hbl), which is continuous into Red Rock Canyon, have ⁴⁰Ar/³⁹Ar biotite ages of 33.78 ± 0.09 and 34.31 ± 0.26 Ma (Lipman et al., 2015), and samples from the andesite-inclusion flow in lower San Juan Creek have biotite ages of 33.67 ± 0.08 Ma and 33.72 ± 0.09 Ma, documenting relatively early eruption of these lavas (Fig. 4; Table 3). A rare sanidine-bearing dacite low along Red Rock Canyon yielded a sanidine age of 35.21 ± 0.03 Ma that is even older. An isolated eastern hill of near-glassy silicic dacite (68.6% SiO₂; ~20% crystals of Pl >> Qz, Sa, Bt), along the margin of the San Luis Valley near the mouth of Sanderson Gulch, has a sanidine age of 34.92 ± 0.02 Ma. The early age at this site must be from the upper part of a thick lava dome deep within the thick tilted volcanic section; this dacite crops out east of much-younger cover by eastward-dipping regional ignimbrites (Fish Canyon, Carpenter Ridge Tuffs).

Considerable uncertainty continues, however, about distinction in some erosional remnants between the lower dacite of Red Rock Canyon versus the upper dacites along the southeastern flank of the BLVC. A related complexity concerns the sizable area of rhyolite (71.6% SiO₂; ~15% Pl > Sa, Bt), exposed for several kilometers beneath andesite lavas between lower San Juan and Cottonwood Creeks, that yielded a sanidine age of 33.68 ± 0.11 Ma. The geographic location and age similarity to other rhyolite lavas exposed low in the central locus assemblage favor including this body as part of the BLVC. Alternatively, elevated concentrations of incompatible elements such as Zr and the light rare earth elements (LREE) suggest closer affinities of this early rhyolite to the Tracy volcano just to the north (Table S2).

In addition to the thick accumulation of relatively old dacite lavas, presence of a discrete northern magmatic locus is indicated by a cluster of five thick dacite dikes (66.0% SiO₂; ~25% Pl >> Hbl > Cpx ± Sa) in the upper reaches of Little Cottonwood Creek. The largest is at least 200 m thick at its widest, is continuously exposed for 1.5 km, and has a regionally atypical E-W strike. These dikes are younger than the dacite lavas of the northern locus; a sanidine age from the large E-W dike is 33.18 ± 0.02 Ma. This age is similar to the less-reliable biotite dates from upper dacitic lavas that flank the BLVC along the eastern mountain front and on the ridge crest to the west (33.4–32.9 Ma). This time equivalence suggests that some of the upper dacite lavas were erupted from late vents within the northern locus.

In addition to the large dacite dikes in this northern part of the BLVC, smaller dikes of rhyolite and andesite, mostly trending NE, intrude the dacite and ande-site lavas. Some, or perhaps all, of these likely are distal dikes from the central locus to the south, or even distal dikes from the Tracy volcano to the north.

Although lavas of the northern and southern loci are older than those in the central area, late dacitic intrusions in both northern and southern sectors are similar in age to the upper dacite lavas that flank the composite centers of the BLVC, emplaced at ca. 33.2 Ma. Such age similarities support treating the three magmatic loci as parts of an overall volcanic complex. Flanks of the BLVC are onlapped by erosional remnants of ignimbrites erupted from central San Juan calderas, particularly the 28.20 Ma Fish Canyon and 27.75 Ma Carpenter Ridge Tuffs. No record of late eruptive activity at the BLVC seemingly remains preserved for the roughly 3.5 m.y. intervening period before the ignimbrite eruptions.

The documented duration of exposed lavas and associated intrusions at the BLVC spans ca. 35–33 Ma, overlapping but largely older than volcanoes to the south. Total exposed topographic relief for these rocks is ~900 m, and the Hellgate drill hole (Fig. 3), on the southwest flank of the BLVC, penetrated 2540 m of Conejos–type rocks in the subsurface (Curtis, 1988). Other drill holes to the south have also documented thick sections of Conejos-like rocks (Table 2). We attempted, unsuccessfully, to date finely ground drill-cutting fragments from zones logged as dacite near the base of the Hellgate hole. Unfortunately, the biotite was too oxidized to analyze, and rare K-feldspars yielded Proterozoic ages, presumably detrital grains from sedimentary horizons.

Despite the substantial subsurface thickness documented by drill holes, the sanidine ages from a dacite in Red Rock Canyon and from the isolated eastern hill of dacite, at the margin of the San Luis Valley, 35.21 ± 0.03 and 34.92 ± 0.02 Ma, respectively, are nearly as old as any dates obtained thus far for inception of volcanism in the San Juan locus (Table 3). This result suggests rapid volcanic growth at the eastern Conejos centers starting ca. 35 Ma.

Another previously little-studied Conejos volcano (Fig. 6), centered ~20 km southwest of the southern intrusive locus of the BLVC, is defined by outward-dipping andesitic to rhyolitic lavas, an intrusive core, and associated dikes in the Baughman-Embargo-Seitz Creek area (Larsen and Cross, 1956; Steven et al., 1974). This eruptive center, designated the Baughman volcano (Lipman, 1976), is located just north of the Rio Grande, where the river has carved a broad valley in laharic volcaniclastic rocks that constitute distal flanks between the Conejos constructs at Platoro to the south and the Baughman, Summer Coon, and other eruptive loci to the north. Our remapping of the Baughman center differs substantially from prior interpretations, especially the locations of central intrusions. Additionally, silicic rocks that were previously described as a discrete “Twin Mountains center” (Larsen and Cross, 1956; Willman, 1993; Parker et al., 2005) are interpreted as the southeastern flank of the Baughman volcano. This edifice appears to have grown nearly concurrently with the much-studied Summer Coon volcano just to the east (Harp and Valentine, 2018, and references therein); the sequence of erupted compositions is closely similar, but details of interfingering boundaries and relative ages remain uncertain.

The bulk of the exposed Baughman volcano consists of proximal breccias and more distal lavas and laharic deposits of finely porphyritic plagioclase-clinopyroxene-(olivine) andesite and basaltic andesite (54–57% SiO₂), similar to that at the more-studied Summer Coon volcano just to the east. Andesites from these two edifices merge indistinguishably, roughly along the east flanks of Twin Mountains and Blue Mountain (Fig. 6; Lipman, 1976). Four samples from a well-stratified sequence of compositionally similar andesite lavas along lower Embargo Creek on the south-dipping flank of Baughman volcano (Fig. 7) have yielded groundmass ages of 32.9–32.3 Ma (Fig. 8; Table S1B). As indicated by the conformable lava sequence, these lavas appear to have been emplaced rapidly. The oldest of these ages (32.86 ± 0.02 Ma) may approximate time of eruption, although uncertainties about variable Ar loss from the fine-grain groundmass suggest that these may be minimum ages.

East- and southeast-dipping lavas of phenocryst-poor rhyolite (72.4% SiO₂; 10% Pl > Bt > Sa) overlie andesite along the east flank of the Baughman edifice, especially on Twin Mountains. Exposures of the rhyolite at variable elevations north of Twin Mountains indicate variable erosion of andesite prior to the rhyolite eruptions, although some northern areas of rhyolite may be at least partly intrusive. A rhyolite at Twin Mountains has a sanidine age of 32.68 ± 0.10 Ma; biotite from two samples at the same site yielded similar but modestly older dates, 33.13 ± 0.11 and 33.08 ± 0.14 Ma. The Twin Mountain sanidine age is close to that for the underlying andesite lavas. A separate sizable rhyolite body that is locally more crystal-rich (72.0% SiO₂; 15%–20% Pl > Bt), just east of lower Baughman Creek, appears to be at least partly intrusive into andesite; it yielded a biotite date of 31.97 ± 0.06 Ma.

Thick lavas of porphyritic dacite (66% SiO₂; ~30% Pl > Bt > Hbl ± Cpx, Sa) overlie rhyolite on Twin Mountains and cap ridges along the northern and northwestern flanks of Baughman volcano (Fig. 6). An additional area of young dacite on the more distal west flank of the edifice has been mapped beyond the area of Figure 6, along the upper reaches of Embargo Creek (Steven et al., 1974). A glassy clast from the basal breccia of a dacite near Groundhog Park on the proximal northwest flank has a robust sanidine age of 32.13 ± 0.03 Ma (Table S1B, sample 19L-13). Biotite and hornblende determinations from this same sample yielded variably older problematic dates, as discussed in the Appendix. Biotite from another nearby dacite body that lacks sanidine yielded a younger date of 31.20 ± 0.01 Ma. The younger dacite has steeply dipping flow layering along its upper sides and the apparent geometry of a deep paleovalley fill with vitrophyre along its base; alternatively, it may be intrusive.

The core of Baughman volcano as mapped by us is an irregular elongate intrusion ~1 × 2 km across, centered on Cyclone Mountain. It mostly consists of aphyric fine-grained dark rocks of andesitic appearance that grade into small areas of fine-grain monzonite (56.5% SiO₂). The andesite differs from nearby lavas in its more densely massive and rectilinear-jointed outcrop appearance, as well as absence of vesicularity, flow brecciation, or any stratification as would be expected between successive flows. Interpretation of the large mass of andesitic rocks on Cyclone Mountain as intrusive would be less convincing, were it not for the presence of similar intergradational associations of intrusive andesite with monzonite in the cores of other early volcanic centers farther north (including Biedell-Lime, Tracy, Jacks Creek), where the larger proportion of coarser rock makes the intrusive character of the associated andesite more obvious. Intrusive andesite from the north ridge of Cyclone Mountain yielded a groundmass age of 31.95 ± 0.06 Ma, consistent with the modestly older ages for lavas on edifice flanks. Unfortunately, attempts to separate zircon from the monzonite at Cyclone Mountain were unsuccessful.

Andesitic and dacitic lavas and smaller intrusions adjacent to the central intrusion have been pyritized and modified by hydrothermal alteration, complicating and obscuring details of intrusive contacts in the core of the Baughman center. The sizable area of altered volcanic rocks, especially northwest of the main mapped intrusive bodies, suggests that a larger composite intrusive complex may be present at depth. These altered lavas and breccias were depicted as a large intrusion by Steven et al. (1974); a second large intrusion, ~2 × 3 km across as shown on that map, is the thick dacite lava flow on the east side of Groundhog Park where we dated sanidine from its glassy basal flow breccia.

Compositionally diverse andesitic dikes, including aphanitic and plagioclase-, augite-, and hornblende-phyric varieties, intrude the lower andesite lavas and breccias but not the overlying rhyolite and dacite. As currently mapped, these small and commonly obscure dikes define an incomplete radial pattern, but with a preferential NW-SE trend (Fig. 6). Groundmass and hornblende ages for three of these dikes range from 32.91 ± 0.07 to 31.74 ± 0.04 Ma, most only slightly younger than for the andesitic lavas.

Rhyolite dikes at Baughman volcano, mostly phenocryst poor (~3–5%) with small plagioclase and biotite, are larger and more laterally extensive than the andesites. NW-trending, crystal-rich and crystal-poor dikes on the northwest flank yielded essentially identical biotite ages (32.53 ± 0.08, 32.57 ± 0.02 Ma), similar to the rhyolite lava on Twin Mountains. The rhyolite dikes align preferentially NW-SE, with more constrained trends than for the andesites.

Dikes and irregular bodies of intrusive dacite porphyry cut the central intrusion and extend into flanking andesite lavas. Duplicate analyses of biotite from a large dacite intrusion that projects westward into Baughman Creek gave consistent ages of 31.14 ± 0.02 Ma and 31.13 ± 0.03 Ma, providing a seemingly reliable indication of late intrusive activity, younger than any identified Baughman lavas. Thick dacite dikes also continue for several kilometers to the northwest and southeast, cutting diverse volcanic units. The NW-trending dacites define the most pronounced linear array of any dike composition at the Baughman center.

The apparent progression at Baughman volcano from more radial dike geometry during early andesitic activity to more linear trends during the late dacites suggests that the geometry of regional stress was transitioning from near-equant to SW-directed extension during main growth of the Baughman volcano within the interval 33–32 Ma. The eruptive sequence, dike geometry, and ages at the Baughman edifice differ from those at the BLVC. Baughman volcano lacks exposed early rhyolite, its andesite mineralogy is anhydrous (no hornblende phenocrysts), the dominant dike alignment is NNW (in contrast to NNE at BLVC), and the age range for main magmatism is more restricted and younger (33–32 Ma). During this main interval of volcano growth, a large-volume ignimbrite (Saguache Creek Tuff) erupted at 32.47 Ma from the North Pass caldera ~30 km to the north, but no remnant of this tuff sheet is preserved in the Baughman–BLVC area. As at the BLVC, the depth to the base of the Baughman edifice and its vertical extent are uncertain, but detailed study of Conejos rocks north of the mapped area (Fig. 6) would likely yield additional information.

The well-exposed Summer Coon stratovolcano is a conspicuous major Conejos eruptive center, onlapping the southern flank of the BLVC and merging to the west with Baughman volcano (Fig. 1). Erosionally dissected remnants of its cone are intruded by spectacularly symmetrical radial dikes (Fig. 9), and both eruptive and intrusive rocks have a broad compositional range from olivine basaltic andesite to rhyolite (52–74% SiO₂). As a result, this volcano has been the subject of far more studies than other parts of the Conejos Formation (Larsen and Cross, 1956; Lipman, 1968, 1976; Mertzman, 1971; Zielinski and Lipman, 1976; Grau, 1989; Perry et al., 2001; Parker et al., 2005; Poland et al., 2004, 2008; Harp and Valentine, 2018). Summer Coon volcano has an eruptive history closely similar to that at the nearby Baughman edifice, but it contains dikes and other intrusions that may be slightly older and displays a much more radial dike geometry.

At exposed levels, Summer Coon contains a cone of outward-dipping basaltic andesite breccias and lavas, overlain by small rhyolite flows and then by thick flanking dacites, completing an upward compositional progression much like that at Baughman volcano. The Summer Coon edifice has been deeply eroded, and cross-sectional reconstruction suggests that its original summit lay 1–2 km above present-day levels (Lipman, 1968, plate 1). The edifice has been erosionally beveled to a surface of modest topographic relief, exposing the radial dikes and a central core of clustered intrusions ranging from andesite to diorite and monzonite (Fig. 9). The central intrusions are surrounded by an outer zone, ~3 km across, of poorly exposed, fine-grained mafic rocks that may be intrusive phases, or alternatively ponded lava within a small axial caldera. Flanks of the eroded edifice are onlapped by ignimbrites (30.2–27.7 Ma) from Platoro and central San Juan calderas.

Despite the excellent exposures and well-defined stratigraphic sequence of Summer Coon volcano, the overall structure and eruptive history of the edifice are incompletely known, in part because of the limited local topographic relief. Upper proximal parts of the volcano that could record late eruptions have been largely eroded, although scattered erosional remnants of Conejos lavas remain preserved on nearby ridges at elevations as much as 500 m above the Summer Coon intrusive core. Petroleum exploration drill-hole logs document an ~1200 m section of subsurface Conejos rocks in the Summer Coon area (Gries, 1985), but cuttings and well logs are inadequate to evaluate how much of this thickness constitutes lower levels within the Summer Coon edifice versus onlapped flanks of adjacent or underlying volcanoes. Andesitic lavas and breccias of the Summer Coon edifice lap out along its north margin against 35.4 Ma hornblende-andesite lava of the BLVC, suggesting that the northern edifice may project beneath Summer Coon at relatively shallow depths.

Several hundred radial dikes have been mapped at Summer Coon volcano (Lipman, 1968, 1974; Mertzman, 1971). Dikes of basaltic andesite are the most abundant and least conspicuous: Subdued in erosional expression, they typically are less than a meter wide and commonly traceable along strike for only a few meters. The andesite dikes intrude only the andesite breccia of the core edifice. In contrast, the rhyolite and dacite dikes form prominent topographic ribs, they are as much as 5–10 m thick, and many are traceable for several kilometers or more. These dikes intrude outer flanks of the Summer Coon edifice as well as the andesite.

Other than the more-radial dike alignment, Summer Coon is closely similar to Baughman volcano just to the east; both contain an early cone of voluminous andesite that lacks hydrous minerals, followed by rhyolite lavas, and then flanking dacites. Lavas from the two volcanoes merge on the slopes west of Old Woman Creek, and boundaries between the two edifices have been difficult to define. In contrast to the preferential NNE dike orientation at the BLVC and NNW direction at Baughman volcano, however, the spectacular radial symmetry for Summer Coon dikes of all compositions, lacking any preferential direction, indicates an isotropic stress field during volcano growth despite proximity to the present-day Rio Grande rift.

Prior to our new results for the BLVC (though a composite volcanic complex, rather than a single edifice), Summer Coon rocks had been dated far more extensively than any other central volcano of the SRMVF. By the early 2000s, 27 samples of Summer Coon lavas and intrusions had been dated by K-Ar, ⁴⁰Ar/³⁹Ar, and U-Pb methods (Table S1C), including multiple mineral phases from several dacites. In addition to early determined K-Ar ages (Lipman et al., 1970), these include eight ⁴⁰Ar/³⁹Ar ages determined at New Mexico Tech (Perry et al., 1999, 2001), our previously unpublished ages for 15 samples analyzed in 2007 with the MAP 215 mass spectrometer, and a U-Pb zircon age for the central monzonite intrusion (Sliwinski et al., 2022). Four ages from sites resampled in 2022 were obtained at higher precision with the Helix MC Plus multicollector mass spectrometer.

These multiple attempts since the late 1960s to determine ages of the successive eruptive and intrusive phases at Summer Coon volcano have yielded inconsistent, and in places contradictory, results (Fig. 10), despite the more widespread presence of presumably suitable minerals (biotite, hornblende) than for most Conejos lavas farther south and west in the San Juan region. Unfortunately, in contrast to BLVC and Baughman volcano, no sanidine-bearing phases have been recognized at Summer Coon volcano. Despite substantial interpretive problems, we summarize available ages for this volcano here, to document sampling and analytical issues, and hopefully to encourage additional high-precision age determinations in the future.

Six plateau and isochron ages, interpreted to be most reliable for the early-erupted andesites of the central cone, range from 33.40 ± 0.25 to 32.90 ± 0.32 Ma (Fig. 10; Table S1C). Three andesite samples yielded much younger plateau ages at 30.48 ± 0.28 to 28.04 ± 0.28 Ma, inconsistent with stratigraphic succession and intrusion ages. These anomalous ages are likely due to loss of ⁴⁰Ar from poorly crystalized fine-grained groundmass, despite careful petrographic screening to avoid samples containing interstitial glass. The preferred andesite ages at Summer Coon appear modestly older than those for early andesites at the adjacent Baughman volcano, although the spread of ages for both suites precludes precise comparisons. The well-stratified conformable successions of compositionally similar andesite breccias and lavas at both volcanoes suggest geologically rapid eruptions and edifice growth. Actual eruption durations at both volcanoes likely were less prolonged than suggested by the range of determined ages, even for the preferred samples.

Most biotite or hornblende phenocrysts in dike rocks yielded older ages than the lavas they intrude, and some biotite-hornblende pairs from the same samples have ages that differed substantially beyond analytical uncertainties (discussed in more detail in the Appendix). Two ⁴⁰Ar/³⁹Ar rhyolite dates (both biotite), from separate sites along the same northeast dike at Summer Coon, differed by nearly a million years. The late flanking dacite lavas (two samples) and associated dikes (10 samples from four sites) yielded especially puzzling dates. Biotite and hornblende pairs from some individual samples of dacite are inconsistent beyond analytical uncertainty, and multiple samples from the same dike yielded divergent ages (Fig. 10).

As one example of inconsistencies, three samples from a western dacite dike (Old Woman Creek) yielded biotite and hornblende dates of 34.01–36.11 ± 0.14 Ma, including a concordant biotite-hornblende pair for the older age. More positively, higher-precision determinations with the Helix mass spectrometer on hornblende from three previously dated sites along the large Natural Arch dike yielded closely similar hornblende ages (33.97 ± 0.07, 33.97 ± 0.08, 33.99 ± 0.06 Ma). One earlier-determined biotite date (33.86 ± 0.09 Ma) from the distal site along La Garita Creek is consistent with the hornblende ages, but another biotite from the distal dike (35.51 ± 0.12 Ma) has an apparent age 1.6 m.y. older.

Additionally perplexing, the dacite lavas and dikes yielded ages older than those from the prior-erupted andesite lavas and breccias. The two samples of dacitic lava have biotite ages of 33.58 ± 0.16 and 33.79 ± 0.12 Ma, slightly older than the earliest dates from the underlying andesites. Most dacite dikes yielded even older ages. Even excluding three of the older dates that are inconsistent with other ages from the same dike or with another paired mineral, the preferred dike ages cluster at 34 Ma, again earlier than those from the intruded andesite lavas and breccias. Either most of the biotite and hornblende dates from the late-emplaced dacite dikes are too old, virtually all groundmass dates from the earlier andesites are much too young, or, most likely, dated phases from both compositional types are problematic to varying degrees. In addition to variable ⁴⁰Ar loss from andesite groundmass and attendant young ages, we increasingly suspect problems from excess argon in biotite and hornblende and attendant anomalously old apparent ages in many of the more silicic rocks. Such problems may be especially common in rapidly quenched samples from near-glassy dike and lava margins (the favorable sites for sampling petrographically pristine phenocrysts), as discussed further in conjunction with data from other Conejos volcanoes (see Appendix).

A U-Pb zircon age of 32.2 ± 0.5 Ma from the central monzonite intrusion (Sliwinski et al., 2022), although lower in precision, plausibly records the youngest magma to crystallize at Summer Coon and is broadly consistent with dates from andesite and rhyolite phases. Thus, although the Summer Coon ages are difficult to evaluate with confidence, a preferred interpretation based on the older andesite groundmass ages is that this volcano was mainly active for ~500 k.y., largely between 33.5 and 32 Ma, nearly concurrent with growth of the Baughman edifice just to the west. This would require that nearly all dates from the dacite dikes are older than the actual time of intrusions. Alternatively, if reliable, the younger mineral ages from dacites would suggest the main activity at ca. 34 Ma and earlier, and an evolution predating Baughman volcano except for much later emplacement of central intrusions. Perhaps contrary to this second alternative is the common good agreement between groundmass and mineral ages from individual samples at other San Juan sites (Table S6).

Scattered low-relief exposures of aphyric and hornblende-bearing andesite and dacite (59.5–67.7% SiO₂), forming low hills over an equant area ~6 km in diameter along the Rio Grande and centered near the town of Del Norte (Fig. 9), probably constitute a small Conejos volcano separate from the Summer Coon edifice (Lipman, 1976; Parker et al., 1991, 2005). The common hornblende-phyric lavas in the Del Norte area are distinct from the anhydrous mineralogy of Summer Coon andesites, and the small crystal size and sparse biotite in Del Norte dacites differ from typical lithologies at Summer Coon. A single hornblende age of 33.85 ± 0.08 Ma (Table S1C) from a Del Norte lava is older than Summer Coon andesites, more comparable to compositionally similar lavas from the southern BLVC.

A thick N-trending dacite dike at Indian Head, a prominent local landmark, yielded a biotite age of 33.18 ± 0.11 Ma, similar to the lava dates deemed most reliable for the Summer Coon edifice. Indian Head may be a late-emplaced feeder for the Del Norte flows, and several NW-trending dikes have been described farther southwest, suggesting an at least partial radial geometry with the volcanic axis concealed beneath widespread surficial deposits along the Rio Grande valley (Parker et al., 1991, their fig. 1). Alternatively, the Indian Head intrusion may constitute a satellitic southern vent of the larger Summer Coon edifice, with associated lavas representing distal south-flank equivalents to upper dacites more proximally at Summer Coon. Although hornblende-rich, the Del Norte rocks define compositions (Table S2C) that, though more limited in range, are similar to Summer Coon pyroxene andesites and dacites at similar stratigraphic levels just to the north (Parker et al., 1991, 2005). The Summer Coon and Del Norte ages that are considered most reliable are ~1 m.y. older than Conejos dikes that were late precursors to the Platoro caldera locus to the south (Fig. 1); they are nearly 3 m.y. older than the Black Mountain Tuff, the first major ignimbrite erupted from Platoro (Fig. 10).

Just north of the BLVC and perhaps interfingering with it, Tracy is a deeply eroded volcano, distinguished by a different areal pattern of dikes and by contrasting magmatic chemistry. Low elevations and sparse vegetation on south-facing slopes provide exceptional exposures, especially for eastern sectors of the volcano. It contains compositionally diverse lavas on its outward-dipping flanks, a core area of thick andesite that appears to have ponded within a caldera ~3 × 4 km across, and an axial intrusion peripheral to which a partial radial swarm of andesitic to rhyolitic dikes is concentrated on the northeast to east flanks (Fig. 11). Map relations, isotopic ages, and rock chemistry have been reported previously (Lipman et al., 2015; Lipman, 2020); growth of the edifice is summarized here.

Just as at the BLVC, the oldest exposed proximal rocks of Tracy volcano appear to be small bodies of crystal-poor silicic rhyolite that underlie andesite lavas. A northeastern rhyolite (76.7% SiO₂; ~15% Pl > Sa > Qz, Bt > Hbl) has a sanidine ⁴⁰Ar/³⁹Ar age of 33.38 ± 0.12 Ma (Fig. 12; Table S1D). Biotite and rare hornblende from this sample yielded perplexingly older ages (33.98 and 34.33 Ma, respectively) that are considered less reliable. Lower dacite lavas that crop out below andesite in Red Rock Canyon along the south volcanic flank may constitute early eruptions from Tracy volcano; more likely, these massive flows are a northern phase of the BLVC (Fig. 3).

Varied andesite lavas constitute the bulk of the exposed Tracy edifice. Finely porphyritic to aphanitic andesite (56–61% SiO₂; to 10% small Pl >> Cpx ± Hbl) is present on all flanks, locally interfingers with dacites, and forms the thick central flows inferred to have ponded within a small caldera. Coarsely porphyritic lava and breccia (56–60% SiO₂; 15–25% Pl >> Cpx), with tabular plagioclase to 1 cm across, form much of the volcano’s north flank.

Above the rhyolite and interlayered with andesite lavas are several thick flows of distinctive light-gray small-phenocryst dacite (66–68% SiO₂; ~5%–10% Pl > Bt) characterized by finely laminated flow layering and parallel alignment of abundant thin bronzy-oxidized biotite flakes. No similar dacites have been recognized at Conejos volcanoes to the south. A biotite date of 33.77 ± 0.09 Ma from the small-phenocryst dacite along Tracy Canyon appears to be too old by at least 0.4 m.y., compared to the sanidine age from underlying rhyolite.

The high west flank of Tracy volcano, cresting at Tracy Mountain (3358 m), consists largely of multiple flows, individually as much as 50 m thick, of gently dipping phenocryst-rich mafic dacite (61–64% SiO₂; 15–25% Pl >> Bt ± Cpx, Hbl). These lavas essentially are a northward continuation of dacite lavas that cap the drainage divide along the west flank of the BLVC. At Sierra de la Lola, outflow Bonanza Tuff (33.35 ± 0.03 Ma) wedges out between lower andesite lavas of the volcano and the thick upper lavas of mafic dacite that define the arcuate west flank at the head of Tracy Canyon and on Tracy Mountain (Lipman, 2020). A biotite date of 33.93 ± 0.10 Ma from the basal dacite on the south slope of Tracy Mountain may be at least 0.6 m.y. too old, based on its projected position above Bonanza Tuff at a lower stratigraphic level to the north. More plausible is the biotite age of 33.23 ± 0.02 Ma from an upper BLVC dacite farther south along the continuation of this ridge (Fig. 3).

Despite the sparse age data (results from only four Tracy samples), the control from the Bonanza Tuff in conjunction with the sanidine age of 33.38 ± 0.12 Ma for the early rhyolite constrain the construction duration tightly—within analytical uncertainty at less than 200 k.y. for the voluminous exposed andesites and crystal-poor dacites at Tracy volcano. An upper andesite that caps Tracy Mountain yielded a distinctly younger groundmass date at 31.78 ± 0.09 Ma. This age may document a late eruption at the Tracy edifice; alternatively, an anomalously young age could be due to argon loss.

The Tracy edifice has an east-west extent of at least 15 km at present erosional levels. Because the volcano is eroded or surrounded by younger deposits, except perhaps on its deeply eroded south side, the exposed extent of the edifice is a minimum for its original basal size. If the lower dacite of Red Rock Canyon to the south is the northern part of the older BLVC to the south, then Tracy volcano may have only a limited thickness below present levels, because paleohills of Proterozoic metamorphic rocks are exposed along the Saguache River valley to the north (Lipman, 2020).

Rhyolite and dacite dikes at Tracy volcano typically are several meters or more thick and make conspicuous outcrops that are traceable for substantial distances. Andesitic dikes are smaller, crop out discontinuously and obscurely, and likely are more numerous than have been mapped. These dikes are especially conspicuous where they intrude the small-phenocryst dacite at sparsely vegetated lower elevations. The confidently mapped absence of andesite dikes within this dacite unit south of Tracy Creek, in contrast to their abundance to the north, documents the asymmetrical radial geometry of dike intrusion at Tracy volcano especially clearly. The elliptical central intrusion, almost 1 km in longest dimension, consists mainly of near-aphanitic intrusive andesite, with a core of texturally variable monzonite. A sample of the monzonite phase collected for U-Pb dating yielded no zircons.

This volcano is near the southern end of the ignimbrite-caldera trend of similar age along the Sawatch Range (Fig. 1), from Grizzly Peak (34.75 Ma) at the north southward to Bonanza (33.35 Ma). The strong eastern concentration of dikes at Tracy volcano, in contrast to the near-radial distribution at Summer Coon volcano, suggests a stress asymmetry involving buttressing by shallow Proterozoic basement to the west, analogous to that modeled for dikes at Spanish Peaks, along the east flank of the Colorado Rockies (Odé, 1957). Proterozoic rocks are exposed along Saguache Creek just north of Tracy volcano and form rugged paleohills farther north, but the basement geometry beneath Tracy volcano and to the west is poorly known. Presence of such basement rocks at shallow depth would be consistent, however, with southward projection of a high-standing paleo–Sawatch uplift. Such a structural interpretation is supported by more clearly exposed relations between dike geometry and exposed Proterozoic paleohills at Jack Creek volcano 15 km to the northwest.

Farther north across the Saguache River valley, another well-defined Conejos-type volcano in the northeastern San Juan region is centered in the Jacks Creek drainage (Lipman, 2020). Jacks Creek volcano is a symmetrical central volcano ~10 km in diameter at present-day outcrop level (Fig. 13); low elevations and limited vegetative cover provide especially fine exposures. The edifice consists mainly of outward-dipping andesitic lavas and laharic breccias that overlie rhyolite near the volcano core. Lavas of fine-grained biotite dacite are discontinuously present along the outer volcano flanks. Axial parts of Jacks Creek volcano were deeply and rapidly eroded during the ~1 m.y. prior to eruption of the Bonanza Tuff that laps depositionally onto the volcano flanks. Some dikes near the core are directly overlain by this ignimbrite.

Radial dikes of andesite to low-silica rhyolite are well exposed on the south and east flanks of the Jacks Creek edifice but are absent to the north and west. A composite central-vent intrusion consists of fine-grained granodiorite, intrusive andesite, and porphyritic rhyolite. The variably altered rhyolite appears in part intrusive but also includes pumiceous breccia that may be remnants of vent fill.

Lavas and intrusions of the Jacks Creek volcano have yielded ⁴⁰Ar/³⁹Ar determinations that are mostly in the range of 35–34 Ma (Fig. 14; Table S1E), among the oldest dated rocks from the San Juan region (Table 3). A large dike of porphyritic low-silica rhyolite has a sanidine age of 34.45 ± 0.10 Ma. The central body of altered porphyry, in part pyroclastic vent breccia, yielded an indistinguishable age of 34.53 ± 0.06 Ma. A U-Pb zircon date from the central altered porphyry agrees closely at 34.5 ± 0.2 Ma (Sliwinski et al., 2022), and the adjacent core intrusion of monzonite has a zircon date of 34.1 ± 0.4 Ma. Attempts to obtain groundmass ages from sanidine-absent lavas and dikes of intermediate composition yielded inconsistent results; most results are in the range 34.5–34.1 Ma, but several groundmass ages of andesite lavas are younger than mineral ages for crosscutting dikes (Fig. 14). Groundmass from the basal rhyolite flow yielded disturbed ⁴⁰Ar/³⁹Ar age spectra: 34.41 ± 0.22 Ma (integrated) and 35.61 ± 0.11 Ma (isochron).

Jacks Creek thus appears to be the oldest well-defined volcanic center within the northeastern San Juan region, predating eruption of the 34.15 Ma Thorn Ranch Tuff from Marshall caldera ~15 km to the north. The north and west flanks of the Jacks Creek edifice bank against paleohills of Proterozoic rocks and are onlapped by younger intermediate-composition lavas that appear to fill southern parts of Marshall caldera. The high-standing basement rocks adjacent to this edifice are interpreted as a southerly continuation of early Cenozoic (Laramide) Rocky Mountain uplift along the Sawatch trend. The asymmetrical distribution of radial dikes at Jacks Creek, absent on the western and northern flanks where basement rocks are exposed, is geometrically comparable to the configuration inferred to be a buttress on dike geometry at Tracy volcano to the south.

Adjacent to the Conejos edifices along the eastern San Juan flank, two large composite accumulations of intermediate-composition lavas to the north and south were the proximal antecedents to voluminous ignimbrite eruptions and attendant caldera collapses that caved away large portions of the precursor volcanoes. To the south at the polycyclic Platoro caldera complex, outer flanks of several earlier Conejos edifices are preserved at caldera walls. To the north, the caldera that formed during eruption of the Bonanza Tuff is similarly geometrically related to a preexisting highland consisting of diverse lavas of the Rawley volcanic complex.

The type locality of the regional Conejos Formation (Conejos Peak, 4017 m) forms the high point along the eroded south rim of the Platoro caldera complex (Fig. 15), from which at least six large dacitic ignimbrites (100–1000 km³) erupted between 30.2 and 28.8 Ma (Lipman, 1975a; Lipman et al., 1996; Lipman and Zimmerer, 2019). Conejos lavas and laharic breccias are exposed over a topographic range of 1150 m from Conejos Peak to the valley floor, with no base exposed. Stratigraphic and compositional relations among Conejos rocks at Platoro caldera and more broadly in the San Juan region document that clustered large volcanoes of intermediate composition became sites for ignimbrite eruptions and caldera collapse. In these respects, Platoro is closely comparable to younger well-known arc volcanoes that sourced caldera-related ignimbrites, such as Crater Lake in Oregon, Aso caldera in Japan, or Santorini in Greece.

Conejos lavas proximal to the Platoro caldera complex (bisected by the valley of the Alamosa River) have long been interpreted as remnant outer flanks of clustered volcanic edifices, the upper parts largely caved away by caldera subsidence (Lipman, 1975a; Colucci et al., 1991; Lipman et al., 1996). Proximal lavas dominate along caldera rims; proportions of interfingering volcaniclastic rocks increase distally. The voluminous Conejos lavas proximal to Platoro caldera have not been mapped in detail, but they are dominantly andesite and mafic dacite (~56%–63% SiO₂) that vary from aphyric to plagioclase-, pyroxene-, and hornblende-phyric. Basaltic and rhyolitic compositions are absent or nearly so, and biotite-bearing silicic dacite is minor. Calc-alkaline hornblende andesites that are common on the northeast flank of Platoro caldera were designated the Horseshoe Mountain member, and more petrographically diverse lavas as the stratigraphically higher Willow Creek member based on relations along the northeastern flanks of Platoro caldera in a pioneering study of Conejos lavas (Colucci et al., 1991).

A distinctive assemblage of porphyritic lavas (platy-plagioclase andesite of Lipman, 1975a), designated the Rock Creek member by Colucci et al. (1991), interfingers between the Horseshoe Mountain and Willow Mountain members along the east side of the caldera. It is characterized by coarsely plagioclase-clinopyroxene–phyric alkalic andesite that is compositionally unique in its high Zr, Nb, and REE. The Rock Creek lavas are spatially associated with a small-volume welded ignimbrite of mafic plagioclase-pyroxene dacite (tuff of Rock Creek; Lipman, 1974, 1975a) that is similarly compositionally distinct in its anhydrous phenocrysts and elevated trace-element compositions (Table S2F). The Rock Creek lavas and associated tuff constitute another of the several eruptive centers clustered within the area that developed into the Platoro caldera complex, but the genesis of its contrasting composition remains poorly understood. No comparably alkalic rocks are known elsewhere within the San Juan volcanic locus.

Andesitic lavas with platy-plagioclase textures are also present on the distal west flank of Platoro and widely elsewhere in the San Juan region, but limited chemical data lack the uniquely alkalic compositions of the Rock Creek lavas. The two calc-alkaline assemblages of the Conejos Formation in the Platoro area are not easily distinguished where Rock Creek units are absent (Colucci et al., 1991), and further stratigraphic complexity and petrologic overlap likely exist within southern and western areas assigned to these two members. Additional sizable masses of Conejos lava that are exposed farther south and west in the San Juan region (Steven et al., 1974; Lipman, 1975b) must have erupted from sites less related to the Platoro locus; these have been mapped only at reconnaissance scales, and much remains to be learned about their compositions and ages.

Except for the small Rock Creek center on the east side of the caldera, all other lavas and dikes analyzed from the Conejos Formation in the Platoro area are compositionally notable in containing lower concentrations of trace elements such as Rb, Ba, Zr, and REE (Table S2F) than Conejos rocks of similar age farther north. These compositional characteristics persisted during subsequent eruptions of ignimbrites and caldera-filling lavas at Platoro (Lipman et al., 1996; Lipman and Zimmerer, 2019; Gilmer et al., 2021).

Important evidence on the edifice geometry, eruptive duration, and regional stress field during growth of the precaldera Conejos volcanoes and the Platoro caldera cycle is provided by the numerous andesite to dacite dikes that radiate westward from Platoro and merge with the Dulce dike swarm that extends southwestward more than 100 km into New Mexico (Lipman, 1974; Lipman and Zimmerer, 2019). All seven dated dacite dikes west of the caldera, the largest in the radial swarm, and several proximal and distal andesite dikes are younger than the last ignimbrite from Platoro caldera. In unanticipated contrast, three dated proximal andesite dikes and three distal western ones, most of these hornblende-bearing, yielded ages of 30.17 ± 0.08–31.31 ± 0.08 Ma (Table S1F), indicating emplacement late during Conejos volcanism and prior to initial ignimbrite eruptions from Platoro. Another distal radial dike, a sanidine dacite 8 km southeast of the caldera, also has a late Conejos age (30.36 ± 0.04 Ma). Without more age determinations, however, the proportions of Conejos versus postcaldera dikes in the radial swarm remain underdetermined, and the two age groups could not be separated on Figure 16.

Both the precaldera and younger dikes converge toward a common locus within the Platoro caldera complex, approximately coincident with the site of the large postcollapse Alamosa River pluton (Fig. 15). Together, these intrusions are interpreted to record the existence of a long-lived upper-crustal magmatic locus at Platoro that was initially established during Conejos volcanism. The radial dike geometry suggests that the vents inferred to have been located within the caldera area were components of an integrated volcanic cluster influenced by a unified radial stress regime. We refer to this composite of eruptive centers as the Alamosa volcanic cluster, reflecting the geographic locus of dike convergence. Rock Creek compositions are absent among the seven dated radial dikes of confirmed Conejos age. This eastern eruptive locus with its regional distinctive compositions appears to have constituted a discrete small volcano that was petrogenetically unrelated to the composite Alamosa volcanic cluster within the caldera. None of the dated granitoid intrusions in the Platoro area (Gilmer et al., 2021) has a precaldera age, but a N-trending andesite dike at the head of Cañon Diablo (Lipman, 1974) that has a groundmass age of 31.81 ± 0.16 Ma (Table S1F) cuts an undated monzonite pluton that may have cored a Conejos center beyond the area of caldera collapse.

An additional geometric complexity is the preferential radial emplacement of dikes westward from the Alamosa locus that has been suggested as indicating influences from the regional tectonic framework (Lipman and Zimmerer, 2019, their fig. 18). The west flank of the broad Laramide-age Rocky Mountain uplift, with its axis roughly coincident with the subsequent Rio Grande rift, may have formed an eastern buttress and barrier to dike emplacement, causing dikes to propagate preferentially westward toward the San Juan Basin. No subsurface control is available for depth and structure of prevolcanic rocks at the Platoro locus, but a paleohill of Proterozoic granite that protrudes though volcanic cover in the Conejos River valley 20 km south of the caldera (Cross and Larsen, 1935; Lipman, 1975b) documents marked uplift relative to the Mesozoic strata west of the Continental Divide. This interpretation of basement buttressing is analogous, though on larger scale, to that inferred for the Jacks Creek and Tracy volcanoes to the north.

Only a few samples from the enormous areal extent of early-intermediate lavas in the southeastern San Juan region have yielded reliable isotopic ages (Table S1F), and much remains to be learned about the overall eruptive history. In most of the relatively rare sites where basal rocks are exposed, the Conejos Formation consists of volcaniclastic rocks or lavas that lack datable minerals. Due west of Platoro along U.S. Highway 160, however, a small road cut in porphyritic dacite that overlies Cretaceous shale has yielded concordant biotite and hornblende ages at 35.47 ± 0.08 and 35.36 ± 0.07 Ma, respectively, which represent the oldest reliable dates thus far obtained from the Conejos Formation anywhere in the central or southern San Juan region. Attempts to date sparse fragments of K-feldspar from an isolated exposure of crystal-poor nonwelded tuff at about the same horizon, the tuff of Snowball Park (Lipman, 2006), yielded only Proterozoic ages. In the same section 425 m higher along Highway 160, a 30.59 ± 0.09 Ma (plagioclase) date from a conglomerate clast suggests as much as 5 m.y. of eruptive activity for this sequence (Lipman and Zimmerer, 2019). This sample was collected because, atypically for the area, it contained sparse thin biotite flakes; unfortunately, this phase could not be separated successfully.

Earlier-published Conejos ages, by less-precise ⁴⁰Ar/³⁹Ar methods from six hornblende-or biotite-phyric lavas in the Platoro area, vary from 34.1 ± 1.0 to 29.7 ± 0.4 Ma (Colucci, 1990; Colucci et al., 1991). All these are from upper stratigraphic levels in the cluster of precaldera volcanic edifices close to caldera margins (Fig. 16A), and notably, the oldest sample, with concordant biotite-hornblende ages at 34.1–33.7 Ma, is from a high point (Pintada Mountain) on the present-day remnant of the caldera rim. Hornblende from a lava near Willow Mountain on the southeast rim has an age of 32.63 ± 0.22 Ma, supporting the inference (Lipman et al., 2022c) that the precaldera Conejos constructs had achieved much of their total size 1.5 to >2 m.y. prior to the initial major ignimbrite eruption of the Black Mountain Tuff at 30.19 ± 0.16 Ma.

In contrast, the seven dated radial Platoro dikes with Conejos ages (31.3–30.2 Ma) indicate later emplacement, some shortly before the initial ignimbrite eruption (Fig. 16A). Several of these dikes were sampled at elevations high in local Conejos sections (to 3650 m), close to levels of nearby basal ignimbrites erupted from Platoro. This geometry seemingly implies rapid erosion of upper parts of precaldera volcanic edifices late during continued volcanism. Rapid synvolcanic erosion during growth of Conejos volcanoes in the Platoro area is also documented by the vast extent and volume of laharic breccias and conglomerates that interfinger with distal Conejos lavas (Lipman, 1974, 1975b, 1976).

A significant result of this erosion of primary constructs and filling of basins with volcaniclastic rocks was production of subdued paleotopography on which successive ignimbrites erupted from the Platoro caldera complex were deposited. All the major Platoro ignimbrites form widespread outflow sheets, without the sizable local variations in thickness that would have indicated deposition in major paleovalleys. The ignimbrites do thin and wedge out against high-standing Conejos lavas on the proximal east and south flanks of the Platoro caldera complex, in accordance with the interpretation that these lavas represent remnant flanks of clustered precaldera volcanic constructs that lay largely within the subsequently subsided area (Lipman, 1975a). Only at a small segment along the northwest caldera margin are ignimbrites preserved above Conejos lavas on a fragment of the original topographic rim (at Prospect Mountain; Lipman, 1974, 1975a, fig. 64).

As at Platoro, Bonanza caldera formed within a preexisting constructional highland consisting mainly of proximal intermediate-composition lavas (Fig. 17). The dominant volumes of andesite (56–61% SiO₂), along with local lavas of porphyritic dacite and rhyolite (62–73% SiO₂) high in the accumulation, define the eroded remnants of a large high-standing cluster of related volcanic centers, parts of which were caved away during caldera collapse. Long designated the Rawley Andesite (Burbank, 1932; Varga and Smith, 1984), this assemblage more recently has been referred to as the Rawley volcanic complex in recognition of its broad compositional range (Lipman et al., 2015; Lipman, 2020).

The Rawley volcanic complex displays a general stratigraphic sequence that becomes more silicic upward: from relatively fine-grained andesite with small plagioclase and hornblende phenocrysts to more porphyritic and silicic lavas. These include dark porphyritic andesite containing large blocky phenocrysts of plagioclase but little or no biotite, small-phenocryst biotite dacite, and then more coarsely porphyritic lighter-colored biotite dacite, and, in places at highest levels, sanidine-bearing dacite or crystal-poor rhyolite. Although Bonanza caldera is transitional in age and location between the San Juan locus and the older caldera alignment along the present-day Sawatch Range (Fig. 1), both the pre-ignimbrite and caldera-fill lavas at Bonanza caldera are equivalent in age to Conejos volcanoes farther south and west.

Growth of the Rawley volcanic complex is stratigraphically constrained between 34.15 Ma (Thorn Ranch Tuff) and 33.35 Ma (Bonanza Tuff), even though only a few relatively silicic lavas could be dated reliably (Table S1G). Much of the Rawley volcanic complex is similar in composition and age to caldera-filling lavas of the adjacent Marshall caldera, and distinctions between the two assemblages are somewhat arbitrary, especially within the buried southeastern parts of Marshall caldera (Fig. 17). Several sanidine-bearing dacites that are stratigraphically high on the west flank of the Rawley volcanic complex have relatively old ages of 33.89 ± 0.09 and 33.78 ± 0.02 Ma, and a dacite plug on the east flank has a sanidine age of 34.13 ± 0.09 Ma. Groundmass and biotite ages for andesite flows, though of lesser reliability than sanidine ages, vary in age from >34 Ma to as young as 33.5 Ma for a small area high on the southeast rim of Bonanza caldera; these ages indicate some continued activity until shortly before eruption of the caldera-related ignimbrite (Fig. 16B).

Uniquely in the SRMVF, complete sections of the pre-ignimbrite volcanic section are exposed within Bonanza caldera due intense resurgent uplift and deep erosion. Early lavas are as much as 550 m thick on the resurgent dome, and the sanidine age from an uppermost dacite flow below the intracaldera Bonanza Tuff is 33.94 ± 0.04 Ma, only modestly younger than that for a thin dacitic ignimbrite (34.04 ± 0.10 Ma, probably Thorn Ranch Tuff) exposed at the base of the volcanic sequence. Thus, the caldera-floor section of pre-Bonanza lavas formed within 0.1–0.2 m.y., with the latest lava emplaced 0.6 m.y. before eruption of the Bonanza Tuff.

Additionally, two rhyolitic lavas high in the Rawley volcanic complex along the southeast wall of Bonanza caldera have sanidine ⁴⁰Ar/³⁹Ar ages of 34.02 ± 0.02 and 34.06 ± 0.02 Ma, i.e., notably older than the Bonanza Tuff. In comparison, two more-distal southern rhyolites that are lower in the Rawley volcanic complex have similar sanidine ages: 34.03 ± 0.07 and 34.10 ± 0.11 Ma. These dates suggest that the Rawley volcanic complex records waning of the Marshall caldera cycle as well as early precursors to eruption of the Bonanza Tuff (Fig. 16B).

Several small granitoid intrusions and scattered dikes are probably associated with the Rawley volcanic complex, but most have not been dated. In addition to the dacite plug on the east flank, a fine-grained monzonite intrusion (61.9% SiO₂) ~0.5 km across at the Klondike mine on the west side of Bonanza caldera (Lipman, 2020) has a U-Pb zircon age of 33.6 ± 0.4 Ma (Table S1G), indicating probable emplacement during Rawley volcanism. Another monzonite body of similar size and composition 5 km to the north likely is another Rawley intrusion; both are associated with strong hydrothermal alteration. Related dikes have not been recognized, in part perhaps due to heavy forest cover and sparse exposures.

Something of a puzzle is a U-Pb zircon weighted mean date of 34.07 ± 0.28 Ma from monzonite of the large Turquoise Mine pluton (3 × 7 km) that intrudes high into the Rawley lavas on the east flank of the elliptical resurgent dome within Bonanza caldera. This pluton and several smaller exposures of monzonite farther west have yielded ⁴⁰Ar/³⁹Ar ages consistently close to or slightly younger than the Bonanza Tuff; these have previously been interpreted as postcollapse intrusions associated with caldera resurgence (Lipman et al., 2015). A new U-Pb zircon weighted–mean date, from one of the western monzonites high in Rawley Creek, is 33.5 ± 0.4 Ma (Table S1G), which is within analytical uncertainty of the ⁴⁰Ar/³⁹Ar age of 33.24 ± 0.08 Ma from the same outcrop. We interpret the older zircons from the Turquoise Mine pluton as antecrysts from the Marshall caldera cycle, rather than crystallization during assembly of this presumed resurgent intrusion. Alternatively, this pluton, although far larger than those at other Conejos-type volcanic centers, perhaps could have been emplaced high into a Rawley volcanic edifice, which was already nearly completely constructed by ~0.6 m.y. before eruption of the Bonanza ignimbrite and caldera collapse.

In addition to the eastern suite of Conejos-type volcanoes, extending more than 100 km from the Alamosa cluster at Platoro northward to the Rawley volcanic complex at Bonanza caldera, some map, age, and compositional data have been obtained recently for several additional scattered centers farther west and north (Tables S1H and S2I). Although of variable quality and quantity, these data provide additional perspectives on the areal distribution, tempo, and compositional variability of the early volcanism prior to large ignimbrite eruptions of the San Juan locus and Sawatch trend. Notably, in comparison to the Conejos Formation farther south and west, where silicic lavas are rare to absent, the northeastern centers locally contain scattered near-basal flows of rhyolite and dacite, some bearing sanidine and yielding reliable ages for inception of San Juan volcanism.

Thick sequences of finely porphyritic andesite transitional to mafic dacite along the slopes of the Continental Divide west of Bonanza caldera (Lipman, 2012, 2020) are interpreted as another large Conejos edifice (Fig. 1). The Sargents Mesa volcano contains early- and late-erupted lavas of phenocryst-poor rhyolite and minor coarsely porphyritic dacite, in addition to the dominant volume of intermediate-composition lavas. Step-heating plateau and integrated ⁴⁰Ar/³⁹Ar ages for groundmass and biotite from three andesite lavas are in the range 33.49 ± 0.21–33.06 ± 0.15 Ma, i.e., analytically indistinguishable from that of Bonanza Tuff. Several of the analyses are characterized by complex spectra and young ages from low-temperature steps, suggesting these lavas may have erupted earlier than suggested by the measured ages. Bonanza Tuff laps out against the north and southeast flanks of Sargents Mesa, confirming accumulation of most lavas on the mesa prior to eruption of this ignimbrite. Inception of the Sargents Mesa volcano, or lavas precursor to it, is constrained by two similar sanidine ages, averaging 34.27 Ma, from a silicic dacite lava that underlies the distal northwest flank of the thick andesite accumulation, close to contacts with Proterozoic basement rocks.

A large variably altered intrusion of porphyritic dacite, which grades into fine-grained granodiorite, is exposed at roof level in the headwaters of Long Branch on the north flank of the Sargents Mesa edifice (Lipman, 2012). Exposures of this intrusion are ~3 km across, with a nearly horizontal upper contact and steeply dipping sides. The geographic location and closely concordant ⁴⁰Ar/³⁹Ar ages of 33.22 ± 0.10 Ma (Bt) and 33.19 ± 0.21 Ma (Hbl) from an unaltered marginal facies suggest that this body is a late central intrusion of Sargents Mesa volcano. Erosional remnants of Bonanza Tuff, preserved on ridges adjacent to the intrusion at similar elevations, indicate rapid unroofing if the dacite is correctly interpreted as the core of Sargents Mesa volcano. In contrast to other eastern Conejos-age volcanoes described here, no dikes were identified in proximity to the Sargents Mesa intrusion, but this area is remote and heavily timbered, and field study has been limited.

The eastern termination of the Sargents Mesa accumulation was initially interpreted as the partly buried west margin of the 34.3 Ma Marshall caldera, but ⁴⁰Ar/³⁹Ar ages demonstrate that eruptions of these lavas postdate the Marshall caldera and were close in time to the Bonanza cycle. Most plausibly from available ages and compositions, at least the late lavas of this inadequately studied volcano erupted from near the west margin of Marshall caldera, as a peripheral precursor to the ignimbrite eruption of the Bonanza cycle.

A monzonite intrusion (56.3% SiO₂) along Grays Creek ~6 km north of the Bonanza caldera (Lipman, 2020) is a likely site for the eruptive source of the deeply eroded lavas of Conejos andesite and dacite preserved along the southern flank of the Sawatch Range (Minor et al., 2019). Biotite and groundmass dates from scattered exposures of these lavas range from 35.5 Ma to 33.6 Ma (Table S1H). The Grays Creek monzonite yielded an even older weighted mean U-Pb zircon date, 36.6 ± 0.5 Ma, with individual crystals as old as 40.7 Ma. Only four of the 30 analyzed zircons gave ages younger than the oldest dated lavas from this area (Table S4A). The dominant population of older Grays Creek zircons appears to record protracted pre-emplacement crystallization of antecrysts, as are likely present in the monzonite of the southern BLVC locus and the resurgent Turquoise Mine intrusion at Bonanza caldera. In none of these samples were the older zircons appreciably resorbed or otherwise different in morphology from those that yielded more plausible intrusion ages.

The small Sky City mining district (Steven and Bieniewski, 1977) is centered on a monzonite intrusion that cores an andesitic Conejos volcano, as exposed on the eastern wall of La Garita caldera (Fig. 1). The intrusion (59.4% SiO₂) has a U-Pb zircon weighted mean date of 32.6 ± 0.5 Ma (Table S1H) more than 4 m.y. prior to the caldera-forming eruption of the Fish Canyon Tuff (Lipman et al., 2022c). Northeast-trending andesitic dikes, intrusive into Conejos lavas, crop out near the Sky City monzonite (Lipman, 2012), but no age or detailed petrologic data have yet been obtained for the dikes or wall-rock lavas associated with this pre-ignimbrite subvolcanic intrusion.

Several concentrations of Conejos-related dikes are shown east of Sky City on the 1:250,000 scale geologic map of Steven et al. (1974). These are beyond the areas of our detailed mapping, but examination of the most prominent site, a north-trending cluster of seven mapped dikes ~5 km north of Figure 6, suggested alternatively that these features are typical steeply dipping pressure ramps at upper levels of a thick dacite lava flow.

Farther northwest, the Needle Creek intrusion (Fig. 1) is the largest early intrusion in the San Juan region (3.5 × 9 km). It is exposed at near roof level and is texturally variable, ranging from nearly aphanitic andesite to fine-grained monzonite. This body in part consists of coalesced NE-trending dikes; sub-horizontal upper contacts in other places suggest partial growth as a laccolithic complex. Several phases of this intrusion and associated dikes have yielded hornblende ⁴⁰Ar/³⁹Ar ages of 34.66 ± 0.21–34.51 ± 0.18 Ma (Table S1H), suggesting that it constitutes the intrusive core of an early Conejos volcanic center for the thick adjacent sequence of andesite to rhyolite lavas. Nearby dacite flows, low in the lava assemblage, have similar sanidine ⁴⁰Ar/³⁹Ar ages of 34.33 ± 0.05 and 34.21 ± 0.08 Ma. Despite the relatively old ages, northeasterly trends of the overall body and many associated dikes document NW-directed extensional strain during assembly of this composite intrusion. Unroofed southern parts of the Needle Creek intrusion are overlain directly by unaltered Fish Canyon Tuff, documenting deep erosion prior to eruption of this ignimbrite at 28.2 Ma.

A cluster of silicic domes and flows centered along upper Barret Creek laps onto the eroded core of the older Needle Creek volcano just to the east (Fig. 1). These rocks range from nearly aphyric rhyolite (1–5% Sa-Pl >> Bt; 73–76% SiO₂) to crystal-rich silicic dacite and low-Si rhyolite (15–35% Pl-Sa > Qz-Bi; 67–74% SiO₂). Individual flows are 50–150 m thick; preserved exposures of the flow field cover ~60 km². The range of laser-fusion ⁴⁰Ar/³⁹Ar ages for sanidine from four flows is 30.05–29.83 Ma; biotite from a fifth flow yielded a slightly older date of 30.28 ± 0.09 Ma (Table S1H).

The rhyolitic dome field at Barret Creek is atypical of Conejos volcanism. It constitutes the only sizable area of silicic volcanism in the San Juan region during the interval between peak activity of the nearby North Pass caldera (Saguache Creek Tuff) at 32.45 Ma and inception of ignimbrite eruptions from Platoro far to the southeast at 30.2 Ma. The dacitic lavas are compositionally and petrographically similar to the voluminous Fish Canyon Tuff, documenting that generation of this regionally distinctive magma type began relatively early in the San Juan region, well north of the area that would become La Garita caldera. (The 34.26 Ma Badger Creek Tuff, erupted from Mount Aetna caldera in the Sawatch Range, is also petrologically similar.) The Barret Creek lavas, which possibly represent an aborted or failed ignimbrite-scale magmatic system, provide an additional bridge in the transition between earlier activity along the Sawatch trend and younger San Juan centers.

A highly altered Conejos-age intrusion of porphyritic biotite-plagioclase dacite, which forms an erosional basin along the southwest margin of Cochetopa Park caldera, is a probable center for the thick andesitic lavas that underlie the 32.5 Ma Saguache Creek Tuff along the Continental Divide between the Cochetopa Park and La Garita calderas (Fig. 1). These altered rocks are surrounded and covered by unaltered 28.4 Ma Sapinero Mesa and 28.2 Ma Fish Canyon Tuffs. Farther west along the eroded north rim of La Garita caldera, high-standing dacite and rhyolite lavas are as young as 30.51 ± 0.38 Ma (K-Ar, biotite; Lipman, 2006). No compositional data are available for the altered intrusive dacite.

The core of another large central volcano that consists dominantly of andesite is marked by an elliptical pluton of porphyritic hornblende andesite-monzonite, named the Sawtooth intrusion (1 × 1.5 km across; Lipman, 2012), and associated smaller plugs and dikes that intrude Conejos lavas and breccias along the northwest margin of the Cochetopa Park caldera (Fig. 1). Clustered NW-trending dikes of similar hornblende andesite cut Conejos rocks on the floor of the caldera. Both groups of hornblende-bearing intrusions appear to be central parts of a large stratocone that is flanked to the northwest and northeast by the lava piles on Sawtooth Mountain and Razor Creek Dome (Lipman, 2012).

The age of this volcano is stratigraphically bracketed only by the underlying Wall Mountain Tuff (37.25 ± 0.08 Ma) and onlapping Sapinero Mesa Tuff (28.41 ± 0.03 Ma). The volcano may be largely or entirely younger than the Saguache Creek Tuff (32.48 ± 0.03 Ma), but field and age relations remain ambiguous. The capping flow on Sawtooth Mountain yielded a hornblende age of 32.22 ± 0.09 Ma, but a dike cutting the Conejos lava sequence low on the south flank of Razor Creek Dome has a hornblende date of 34.85 ± 0.16 Ma (Table S1H) that likely is anomalously old. Erosional remnants of Saguache Creek Tuff rest directly on prevolcanic rocks around the northwest flank of Sawtooth Mountain, and poor exposures of this tuff on heavily timbered north slopes appear to be overlain by andesite lava flows. In this case, the Sawtooth lavas would have been emplaced during the interval 32.5–32.2 Ma. Alternatively, this edifice might predate and have been severely eroded prior to eruption of the Saguache Creek Tuff that banked unconformably against its northwest flank (Lipman, 2012).

At least some lava in this area erupted earlier: A rhyolite flow banked against Proterozoic rocks in a small area at the northwest margin of Cochetopa Park (Lipman, 2012) yielded a robust sanidine age of 37.38 ± 0.09 Ma, the earliest dated lava from anywhere within the San Juan volcanic locus. This lava overlies the near-aphyric nonwelded rhyolitic tuff of Home Gulch, another local unit exposed only proximal to the Sawtooth center, but attempts to date sparse small fragments of K-feldspar from this tuff yielded only Proterozoic ages.

Multi-hectometer exposures through sections of dominantly intermediate-composition lavas and laharic breccias that underlie the regional ignimbrites are widely and spectacularly exposed in the western San Juan Mountains (Steven et al., 1974; Steven and Hail, 1989), but reliable published age and compositional data are sparse. Early-determined K-Ar ages on biotite and hornblende from western correlative lavas of the Conejos Formation (locally long-designated as the San Juan Formation; Cross and Larsen, 1935) are 33–32 Ma (Table S1I). Several recent U-Pb zircon dates from western lavas and breccias (Gonzales et al., 2021) are in the range 33–30 Ma, together suggesting that widespread volcanism may have begun later in the western San Juan Mountains than at the eastern Conejos volcanoes (ca. 35 Ma). No ⁴⁰Ar/³⁹Ar age determinations appear to have been made thus far for any pre-ignimbrite lavas, probably in part reflecting sampling challenges resulting from widespread propylitic and acid-sulfate alteration proximal to the mineralized western San Juan calderas that have been the focus of most geologic studies. Distal sections through early lavas and breccias along northwest and north flanks of the preserved volcanic extent could offer opportunities for more reliable age determinations.

Several monzonite plutons and associated dikes, intrusive into the thick lavas and breccias of the western Conejos Formation, are possible cores of pre-ignimbrite volcanoes (Steven and Hail, 1989), although limited age and petrologic data have been published. These include the Larsen and Cimarron centers northeast of Lake City, the Matterhorn intrusive locus north of Lake City (Lipman, 1976), the Cow Creek center farther west (Luedke, 1972), and the West Elk volcano to the north (Gaskill et al., 1981; Mutschler et al., 1981). A U-Pb date of 30.2 ± 0.1 Ma on the Calliope dike (Gonzales et al., 2021), which radiates west from the Cow Creek center, appears to confirm a Conejos age for late activity at this center; this is the sole isotopic age control available for any of the western San Juan intrusions. A Conejos age for the outward-dipping lavas and breccias and associated radial dikes of the West Elk volcano is demonstrated by onlapping of the ignimbrites from western San Juan calderas, and several nearby granodiorite plutons have yielded ⁴⁰Ar/³⁹Ar and U-Pb ages in the range 34.6–33.2 Ma (Garcia, 2011). Dikes that radiate outward from the Cimmaron and Matterhorn centers also appear to project beneath ridge-capping regional ignimbrites, but no isotopic ages are available.

The Larsen center, north of Lake City, is defined mainly by massive andesite breccias and thin interleaved lavas, interpreted as remnants of a composite volcano, which are intruded by an elongate (2 km) body of monzonite that has complexly irregular contacts where exposed at roof level on rugged topography (Lipman, 1976). Dikes of andesite and rhyolite extend at least 4 km north and northeast from the monzonite. An inadequately documented hornblende K-Ar age of 32.6 Ma (Lipman et al., 1973) for lavas (Lake Fork Formation of Larsen and Cross, 1956) on the northeast flank of the Larsen center (Table S1I) suggests Conejos-age activity at this site.

Several other western intrusions, at times previously considered candidates for cores of Conejos centers, have yielded relatively young U-Pb ages. A small monzonitic stock that lacks associated dikes intrudes Conejos rocks at Carson Camp in the Lake City area (Lipman, 1976); it has a U-Pb zircon date of 28.75 ± 0.31 Ma (Sliwinski et al., 2022), which is close to inception of the western ignimbrite flare-up. A nearby stock at Castle Creek that intrudes fill of the Uncompahgre caldera, source of the 28.41 Ma Sapinero Mesa Tuff, yielded a zircon date (29.00 ± 0.11 Ma; Sliwinski et al., 2022) that is anomalously old, probably due to the presence of antecrysts. These intrusions thus become additions to a suite of other documented postcaldera intrusions in the western San Juan region, including Sultan Mountain, Ophir, Capital City, and multiple plutons and laccolithic bodies in the San Miguel Mountains (Bove et al., 2001; Gonzales, 2015). The relation of these intermediate-composition intrusions, and others farther east that lack associated dikes (e.g., Blue Creek, Grays Creek; Table 1), to volcanic constructs is unconstrained.

To the northeast, along the Sawatch magmatic trend, precaldera volcanism appears to have been more limited than at the San Juan locus (McIntosh and Chapin, 2004), although deep erosion has severely limited the surviving record of activity. Remnants of the large Guffey volcano (ca. 36 Ma) and other early lavas are preserved in the Thirtynine Mile area (Epis and Chapin, 1968; Wobus et al., 1990), east of the present-day Rio Grande rift (Fig. 1). In contrast, only small areas of the early volcanics are exposed in proximity to Sawatch-trend calderas, and their outflow ignimbrites (37.2–34.4 Ma) are preserved mainly where deposited in eastern paleovalleys, in places directly on Proterozoic basement or other prevolcanic strata (Chapin and Lowell, 1979; Taylor et al., 1975a, 1975b).

In proximity to the deeply eroded source of the Wall Mountain Tuff (37.19 ± 0.03 Ma), lava of the Calico Mountain Andesite (Dings and Robinson, 1957; Toulmin and Hammarstrom, 1990) has yielded a U-Pb zircon date of 36.76 ± 0.12 Ma (Sliwinski et al., 2022), indicative of a pos-ignimbrite eruption age. Contrary to prior interpretations (Shannon, 1988; Toulmin and Hammarstrom, 1990), this age also supports the inference that the Calico Mountain Andesite was intruded by the ca. 35.5 Ma Princeton batholith, rather than deposited unconformably on eroded granitoid rocks. A prebatholith age for the andesite was initially inferred by us, based on geologic contacts mapped by others (Shannon, 1988; Toulmin and Hammarstrom, 1990), especially relations to present-day topography, which indicate a steep contact between the intrusion against the andesite and parallelism of this contact with textural zones within the batholith.

On the southwest side of the Princeton batholith, a small area of preserved lavas, the Hancock dacite of Shannon (1988), is directly overlain by intracaldera Badger Creek Tuff (34.26 ± 0.06 Ma; Zimmerer and McIntosh, 2012). These lavas yielded a U-Pb zircon date of 34.56 ± 0.09 Ma (Sliwinski et al., 2022), just older than the overlying ignimbrite. The Hancock lavas appear to have been deposited unconformably on granodiorite of the 35.5 Ma Princeton batholith (Shannon, 1988) with conglomerates preserved locally along the contact (our field observations). This relationship documents deep erosion within only about a million years after batholith emplacement.

Farther north, a limited surviving record of early-precaldera magmatism in the vicinity of the 34.8 Ma Grizzly Peak caldera is provided by a thick sequence of andesitic lava at Buffalo Buttes, including an interleaved 38.5 Ma tuff (McIntosh and Chapin, 2004), by the late phase of the composite Twin Lake pluton at ca. 42–40 Ma (Feldman, 2010), and by the 36.4 Ma age for the Middle Mountain porphyry in the same area (Rosera et al., 2021).

Despite varied phenocryst assemblages, almost all early volcanic rocks of the San Juan locus have high-K calc-alkaline compositions, typical of continental-arc volcanism. They define distinct areal trends, becoming modestly more alkalic and enriched in trace elements such as LREE, Zr, Nb, and Th from southern and western San Juan centers northeastward into central Colorado (Fig. 18A).

The large range in bulk BLVC compositions from basaltic andesite to rhyolite (55%–73% SiO₂), despite diverse phenocryst assemblages including hornblende-plagioclase andesite and sparse sanidine-quartz rhyolite, forms fairly coherent trends on silica-variation and trace-element diagrams. No substantial differences in trends seem to be evident among the 35 analyzed lavas and intrusions from the south, central, and north loci (Table S2A). Their compositional ranges are similar to those of younger ignimbrites and lavas at the central caldera complex to the west (Lipman, 2006). The BLVC trends provide a framework for comparisons and contrasts with other pre-ignimbrite eruptive centers farther south and north along the eastern San Juan Mountains.

The relatively few analyzed samples at this volcano, varying from basaltic andesite to rhyolite (Table S2B), have compositions generally similar to those from the BLVC. In contrast to the hornblende-plagioclase andesites at the adjacent southern locus of the BLVC, however, andesites at Baughman volcano contain only sparse small phenocrysts of plagioclase and clinopyroxene; some are basaltic andesite with variably altered olivine. Rhyolite dikes are crystal poor, containing only sparse plagioclase and biotite, but the rhyolite lavas that cap Twin Mountains on the southeast flank of this edifice are the most southern identified Conejos rocks that contain sanidine phenocrysts. A few samples contain relatively high alkalis and incompatible elements (>~275 ppm Zr, >~60 ppm La), which are more comparable to typical compositions from early volcanoes farther to the north (Tracy, Jacks Creek).

Mineral assemblages for the broad range of Summer Coon volcano compositions are similar to Baughman volcano but distinct from those at the BLVC just to the north. Andesites are anhydrous, lacking hornblende phenocrysts. A few olivine-rich dikes are silicic-alkalic basalt (SiO₂ as low as 50.8%). These are the most mafic compositions known from any San Juan volcanic center; rare small lamprophyric intrusions in the western San Juan region (Lake and Farmer, 2015) lack volcanic counterparts and are postignimbrite or have uncertain ages. While the dacites at Summer Coon are crystal rich, containing biotite, hornblende, and clinopyroxene, in addition to plagioclase, sanidine is absent. The rhyolites contain only a few percent of small plagioclase and biotite crystals; sanidine and quartz are absent. Despite these differences in mineral assemblages between the Summer Coon volcano and the BLVC, chemical trends are closely similar (Table S2C). Although hornblende-rich, the Del Norte rocks define a limited range of compositions (Table S2C) that are similar to Summer Coon pyroxene andesites and dacites at similar stratigraphic levels just to the north (Parker et al., 1991, 2005).

The diverse lavas and intrusions at Tracy volcano have major-oxide compositional ranges (e.g., 56.5–76.4% SiO₂) broadly comparable to the Conejos edifices just to the south (Table S2D) but notably higher values of incompatible trace elements such as Zr, Nb, Ba, and LREE, as plotted on variation diagrams (Fig. 18A). In these respects, Tracy volcano marks the southern part of a chemical subprovince in the northeastern San Juan region that continues northward into eruptive centers along the Sawatch trend (Fig. 1).

Although older than Tracy volcano, Jacks Creek rocks have similar chemical characteristics, containing high values of Zr and LREE (Table S2E). These are consistent with location of this edifice at the southern extent of the Sawatch trend.

Farther to the south along the eastern San Juan flank, mineral assemblages and chemical trends for most intermediate-composition Conejos rocks from the Platoro area are distinct from those from Summer Coon and other more northerly Conejos volcanoes (Table S2F). The Horseshoe Mountain and Willow Mountain andesites at Platoro typically have a hydrous hornblende-plagioclase assemblage, in contrast to the sparse small phenocrysts of plagioclase and clinopyroxene, some with variably altered olivine, in Summer Coon and Baughman andesites. While some overlap is present, Platoro andesites tend to be even lower in Ba, Sr, Zr, LREE, and La/Yb than for these volcanoes at comparable silica content (Fig. 18A). The incipient trends of northward-increasing incompatible elements, from the main calc-alkaline lava units from Platoro to Summer Coon volcano, are continued by compositions of the Conejos volcanoes along southern parts of the Sawatch trend farther to the north (Fig. 1).

Notably, although lavas of the Horseshoe Mountain member on the northeast flank of the composite Alamosa volcanic cluster occupy a comparable stratigraphic level within the Conejos Formation as petrographically similar hornblende-rich silicic andesites and mafic dacites of the Del Norte volcano nearby to the north, the Horseshoe Mountain and also the stratigraphically higher Willow Creek lavas at Platoro have notably lower Ba and modestly higher Sr and Y contents at similar silica values. In these respects, the Platoro lavas differ from all the Conejos eruptive centers north of the Rio Grande Valley.

In contrast, the peculiarly alkalic Rock Creek lavas (poorly understood, not plotted on Fig. 18A), which crop out on the northeast flank of the caldera (Colucci et al., 1991), are higher than Summer Coon rocks in Zr and REE, and they are distinctively lower in Ba and Sr. A limitation, however, is the relatively restricted areal sampling and limited resulting number of available rock analyses for lavas precursor to the Platoro ignimbrite flare-up, especially for the large areas exposed south and west of the caldera.

Lavas and intrusions of the Rawley volcanic complex continue the northward trend shown by Tracy and Jacks Creek volcanoes, to compositions that are modestly more alkalic, with further elevated contents of total alkalis, Zr, and LREE than more southerly Conejos volcanoes such as the BLVC, Baughman volcano, and Summer Coon volcano, or precursors to the Platoro caldera complex (Fig. 18A). The compositionally variable Bonanza Tuff, especially its dacitic components, has even more elevated contents of the same elements. Spatially and temporally associated postsubsidence lavas that fill Marshall and Bonanza calderas (Table S1G; not plotted on Fig. 18A) also have compositions similar to rocks of the Rawley volcanic complex.

Some analyzed samples from other northeastern centers close to Bonanza, especially Grays Creek and Sargents Mesa, have elevated Zr and LREE contents comparable to those at the Rawley, Jacks Creek, and Tracy volcanoes, while more western sites have lower values of these elements, similar to Conejos volcanos farther to the south.

Accessible compositional data have been sparse for pre-ignimbrite volcanism in the western region. Only four major-oxide chemical analyses seem to have been published for the early lavas (Larsen and Cross, 1956, their table 21, no. 6; Luedke, 1972, table 1, no. 1; Luedke and Burbank, 2000, their table 1, no. 4–5). However, the U.S. Geological Survey (USGS) National Geochemical Database (https://www.usgs.gov/centers/geology,-geophysics,-and-geochemistry-science-center/science/national-geochemical-database) contains ~60 analyses for lavas and volcanic breccias assigned to the San Juan Tuff/Formation or equivalent early volcanics, many with X-ray fluorescence (XRF) or inductively coupled plasma–mass spectrometry (ICP-MS) trace-element determinations. Some analyses in the database were from highly altered rocks, intended for ore-deposit studies; these were filtered from the present compilation (Table S2J), based on submittal notes, incomplete determinations of major elements, or low totals (<99%). Additional analyses with obviously modified alkali ratios (K₂O >> Na₂O; Na₂O < 2%) were also omitted, as were analyses that lacked trace-element data or were obtained by low-precision emission-spectrographic methods.

The remaining analyses represent 28 samples from early lavas and breccia clasts. About half of these are from large lava blocks, derived from caldera walls, enclosed in the landside-breccia facies of the intracaldera Sapinero Mesa Tuff (Picayune Megabreccia Member). Despite varying degrees of propylitic alteration, reflected by secondary volatile contents (H₂O, CO₂) as high as 6% in some samples, the screened analyses have major- and trace-element compositions closely similar to Conejos lavas and intrusions from central sectors of the eastern San Juan Mountains (e.g., BLVC, Baughman volcano, and Summer Coon volcano). For example, the early-erupted western San Juan volcanic rocks have higher concentrations of La and Zr than volcanic precursors to the southeasterly Platoro caldera complex (Fig. 18B) but lower concentrations than the Conejos volcanoes farther north (Tracy, Jacks Creek, Rawley). More numerous analyses (80) of generally less-altered postignimbrite lavas of intermediate composition (Silverton volcanics), which filled the western Uncompahgre and San Juan calderas within ~0.25 m.y. after eruption of the Sapinero Mesa Tuff (before the Fish Canyon Tuff at 28.20 Ma), have compositions similar to the earlier Conejos/San Juan Formation, further confirming the regional geochemical trends.

These summaries of stratigraphic, age, and compositional features for Conejos volcanoes of the San Juan locus provide an augmented framework for interpreting the overall eruptive history of pre-ignimbrite magmatism. Documented are the widespread inception of mid-Cenozoic volcanism at ca. 35 Ma and the focusing of peak magma supply at individual volcanoes as much as several million years prior to subsequent ignimbrite eruptions. Additionally, varied dike orientations are inferred to record changing regimes of tectonic strain during the early volcanism, and regional and temporal variations in magma compositions provide insights about subvolcanic batholith growth and possible associated crustal thickening.

Age control for the widely distributed Conejos volcanoes remains inadequate; most available data come from the northeastern San Juan region, which has been the focus of recent studies summarized here. Significant age ambiguities arise from the need to date less-than-ideal phases (especially biotite or groundmass concentrate) for many samples, where the resulting “geologic” uncertainties are substantially larger than those from analytical precision. Additional limitations are disparate data from center to center, along with the limitations of field observations and sampling that incompletely represent areal extent, vertical range, or compositional volumes. Nevertheless, along with scattered ages from adjacent areas, broad constraints are emerging for the early history of SRMVF magmatism.

Data now seem adequate to show that initial mid-Cenozoic magmatism became younger southward from the Sawatch trend to the San Juan locus and into New Mexico. The oldest ages (40–36 Ma) are from the north end of the Sawatch Range and vicinity: Buffalo Peaks, Guffey volcano, Calico Mountain Andesite, the mid-Cenozoic phase of the Twin Lakes pluton, and Middle Mountain porphyry (McIntosh and Chapin, 2004; Rosera et al., 2021). Near-basal lavas have ages of 35–34.5 Ma, including several from sanidine, at multiple sites in the northeastern San Juan Mountains, including Jacks Creek volcano, the Needles Creek center to the west, and northern and southern loci of the BLVC (Table 3). To the south, concordant biotite-hornblende ages, at 35.4 Ma from dacite at Treasure Falls, suggest that inception of volcanism at the San Juan locus may have been nearly concurrent over a broad region, but data are lacking for initial lavas anywhere to the west or farther south in the San Juan Mountains. Although near-basal exposures are rare, and many lower lavas of the Conejos Formation lack phenocrysts suitable for age determinations, focused efforts to obtain ages from additional sites have yet to be attempted. Farther southeast, the oldest dated Conejos-equivalent lavas are even younger than at the San Juan locus, ca. 30 Ma in the San Luis Hills in southernmost Colorado (Fig. 1; Thompson and Machette, 1989) and ca. 29 Ma at the Questa-Latir locus in northern New Mexico (Lipman et al., 1986; Zimmerer and McIntosh, 2012).

The areal distribution and volumes of early volcanoes appear to have been highly variable, although much of the eruptive record is concealed beneath younger ignimbrites or has been obscured by erosion. In the northeastern San Juan region, many exposed sequences of Conejos lavas are hundreds of meters thick, and drill holes document multi-kilometer sections (Table 2), but only a few kilometers distant, ignimbrites such as the Bonanza Tuff directly lap onto Proterozoic rocks. Distal accumulations of debris-avalanche, laharic, and conglomeratic detritus derived from the eruptive centers also form massive sections up to a kilometer thick, for example, west of the Platoro caldera complex and widely in western San Juan regions. Such thick volcaniclastic accumulations document rapid synvolcanic erosion.

Many Conejos eruptive centers, as marked by thick proximal accumulations or directly by central intrusions and dikes, tend to cluster near sites of subsequent ignimbrite eruptions and calderas (Fig. 1). Such clustering of precursor volcanoes at the San Juan locus has long been interpreted as magmatic focusing associated with incipient assembly of upper-crustal magma bodies that became sources for the ignimbrites (Lipman et al., 1978; Lipman, 1984). Additional Conejos lavas, including some not clearly related to any exposed central edifice, erupted more distal to the ignimbrite calderas. Examples include dispersed thick sequences of lavas near the Colorado border south of Platoro (Steven et al., 1974; Lipman, 1975b) and continuing into the Tusas Mountains in northern New Mexico (Manley et al., 1987), on the San Luis Hills horst within the Rio Grande rift to the east (Thompson and Machette, 1989), and along the northwest flank of the San Juan Mountains (Steven and Hail, 1989).

North of the San Juan locus, the mid-Cenozoic volcanic cover has been widely eroded, and the extent of volcano growth precursor to ignimbrite eruptions is less determinable. Deposition of outflow ignimbrites directly on Mesozoic or older rocks in some areas suggests that the early eruptions proximal to the Sawatch trend calderas may have been less intense and voluminous than to the south, although such distributions may mainly record more intense synvolcanic erosion, as is also suggested by the limited discontinuous preservation of the northern ignimbrites. More distant from ignimbrite eruptive sites, a few early volcanoes, such as West Elk, Guffey, and sites along the west flank of the Wet Mountains (Fig. 1), are well preserved, and others may have been associated with scattered mid-Cenozoic intrusions (Mutschler et al., 1981; Rosera et al., 2021) where any associated volcanics have been eroded.

Potentially significant for eruptive-hazard assessment and modeling of petrogenesis in continental arcs is the timing of peak precursor volcanism relative to a subsequent ignimbrite eruption (Table 1). Histogram plots of ages for Conejos volcanoes provide a simple approximation of times of peak growth (Fig. 19). The histogram ages tend to decrease southward, especially for latest magmatism, but most of the eastern centers discussed here show a peak at ca. 34.5–33 Ma, well before eruption of areally associated ignimbrites. Jacks Creek at ca. 34.5 Ma and the Rawley volcanic complex at 34 Ma are the oldest, and those near Platoro caldera are the youngest (though with a less-defined peak).

Expectably, intrusions tend to be younger than lavas, though this relation is somewhat obscured for the BLVC histogram by grouping of the three loci. Results for Summer Coon volcano are less reliable because of stratigraphic/geochro nologic inconsistencies between dike dates (mostly older) and intruded lavas (mostly younger). Such histograms could be biased by preferential selection of samples best suited for dating. To the degree that such samples, containing sanidine, biotite, or hornblende, tend to be silicic and late erupted compared to the volumetrically dominant andesite and mafic dacite, their ages should effectively record the peak and later stages of volcano growth.

In addition to simple histograms of ages, peak growth at the early volcanoes that were precursors to ignimbrite eruptions, such as those in the San Juan locus, can alternatively be estimated from ages of the topographically highest remnants of the edifice, while termination of volcano growth can be bracketed by ages of overlying ignimbrites. For example, at the large La Garita caldera, such stratigraphic constraints show that proximal Conejos edifices on caldera flanks were already near their final size by 32–29 Ma, as much as several million years prior to the ignimbrite eruption at 28.2 Ma (Lipman et al., 2022c).

Even though interpretive uncertainties for biotite, hornblende, and groundmass ages, along with the scarcity of datable sanidine, limit our capacity to determine detailed growth histories for most Conejos volcanoes, the new data suggest prolonged magmatism and composite volcanic growth at some centers, spanning as much as several million years (Table 1). For example, at the northern locus of the BLVC, voluminous dacitic lavas erupted as early as 35.2 Ma are intruded by dacites as young as 33.2 Ma (both sanidine ages; Table S1A). At the southern locus, andesites as old as 34.8 Ma (hornblende age) are overlain by anhydrous andesites and then by dacitic lavas erupted at ca. 33.1 Ma (biotite). At Baughman volcano, the andesitic flows (ca. 33 Ma) that form lower parts of the volcano are overlain by dacitic lava erupted at 32.1 Ma and intruded by dacite at 31.1 Ma (Table S1B). In the vicinity of Bonanza caldera, early intermediate-composition lavas are as old as 35 Ma and probably peaked ca. 33.9–33.7 Ma, but continued until just before eruption of the Bonanza Tuff at 33.35 Ma (Fig. 16B). Several other early SRMVF volcanoes (Jacks Creek, Summer Coon) also have yielded sizable spreads of ages, but uncertainties about some biotite and hornblende ages make interpretations of magmatic duration less confident (see discussion in Appendix).

At the Platoro caldera complex, the interval of early lava eruptions is less constrained, but the overall duration of focused magmatism was longer. Converging radial dikes document sustained arc-type andesitic and dacitic magmatism at a common locus within the caldera area from at least as early as 32 Ma to as young as 25.5 Ma (Lipman and Zimmerer, 2019). The same locus of radial dike emplacement continued to be active to ca. 20 Ma, during the transition to bimodal magmatism concurrent with early extension along the Rio Grande rift. Other SRMVF caldera systems also had prolonged magmatic histories of 5–10 m.y. duration, defined by early growth of central volcanoes, culminating ignimbrite eruption and caldera subsidence, and then continued postcollapse magmatism (Lipman, 2007, table 1).

From another perspective, the clustered seven ignimbrite eruptions and associated lavas erupted from within La Garita caldera could collectively be considered records of large-volume postcollapse volcanism that continued for more than 3 m.y. after the huge eruption of the Fish Canyon Tuff, recording a total span, together with the precaldera volcanism, of ~10 m.y. of eruptions from 35 to 25 Ma. In the western San Juan region, precursor lava eruptions began by ca. 33 Ma (Table S1I), ignimbrites and related caldera systems were active intermittently 29–23 Ma, and emplacement of late intrusions continued to as young as 15–10 Ma (Gonzales, 2015). Comparable long-duration focusing of magmatic activity in continental-arc systems has also been documented in the Andes, for example, 11 m.y of eruptive history at the Aucanquilcha volcanic cluster in northern Chile, though without associated ignimbrite eruptions (Grunder et al., 2008). In contrast, postcollapse eruptions and exposed intrusions at some San Juan calderas largely or entirely terminated within ~200 k.y. after the last ignimbrite eruption; examples include Bonanza, Creede, and Lake City (Lipman et al., 2015; Bove et al., 2001).

While substantial uncertainties remain for growth histories of the early eastern San Juan volcanoes, several of the major ignimbrite eruptions and associated caldera subsidence appear to have occurred as much as several million years later than peak growth at areally associated precursor volcanoes (Lipman et al., 2022c). Such lengthy precursor intervals of relative volcanic calm are inferred to have been concurrent with enlargement of the upper-crustal magma reservoir, followed by runaway fractionation to more silicic compositions and rapid generation of eruptible magma capable of producing large ignimbrite eruptions (Lipman and Bachmann, 2015; Sparks et al., 2022; Elms et al., 2023; Smithies et al., 2023).

For example, ages of ca. 34 Ma for a Conejos lava at Pintada Mountain on the northeast rim of the Platoro caldera complex and 32.6 Ma at Willow Mountain on the south rim (Fig. 16A) indicate that the composite pre-ignimbrite volcanic edifice was already high standing as much as several million years prior to initial eruption of a major ignimbrite (Black Mountain Tuff at 30.2 Ma). Similarly, though for a shorter time duration, robust sanidine ages from uppermost dacitic and rhyolitic lavas of the precursor Rawley volcanic complex are in the range 34.1–33.8 Ma, i.e., several hundred thousand years or more prior to eruption of the Bonanza Tuff at 33.35 Ma. Ages for high-standing andesite lavas on the southeast caldera rim, as young as 33.5 Ma, suggest continued modest-volume volcanism closer in time to the ignimbrite eruption, though these biotite and groundmass ages may be less reliable. In contrast, sanidine ages from late-erupted precaldera rhyolite lavas beneath the Bonanza Tuff within the subsided caldera floor are substantially older (34.1–34.0 Ma), indicating that much of the central area that subsided during caldera formation was already a largely formed volcanic construct 0.7 m.y. prior to the ignimbrite eruption.

Perhaps especially convincing are available constraints on timing of the precursor volcanism around the margins of the enormous La Garita caldera associated with the 28.2 Ma eruption of the Fish Canyon Tuff (Table 4). As documented in this article, all preserved lavas and late intrusions at the Conejos volcanoes east of La Garita caldera have ages several million years older than the Fish Canyon Tuff. These include upper dacitic lavas (33.1 Ma) and late intrusive dacite (31.1 Ma) at Baughman volcano, high-standing dacitic lavas and late intrusions (33.2 Ma) of the BLVC, and the monzonite intrusion at Sky City (32.6 Ma) along the east wall of La Garita caldera. A few scattered biotite and hornblende ages from dacitic lavas at other margins of this caldera yielded similar ages.

Ages of overlying ignimbrites provide additional evidence for timing of Conejos edifice growth. Along the northeast rim of La Garita caldera, Conejos lavas are overlain by the 32.4 Ma Saguache Creek Tuff, and the 30.2 Ma Black Mountain Tuff overlies Conejos rocks along the southeast margin. To the west, the overlying Ute Ridge and Blue Mesa Tuffs are younger, 29.0 Ma and 28.65 Ma, but still substantially older than the Fish Canyon Tuff. Of course, some late Conejos-type lavas likely erupted within the 2000 km² area of La Garita subsidence, but the areally minor extent of such lavas interfingered with proximal ignimbrites suggests that such late eruptions were volumetrically minor. The only such preserved unit, the Sheep Mountain Andesite, consists of a few thin lavas between the Masonic Park and Chiquito Peak Tuffs near Wolf Creek Pass near the southeast margin of La Garita caldera (Lipman, 2006). Collectively, these age and stratigraphic constraints suggest that much of the growth of the Conejos volcanoes in the area that became the La Garita caldera was largely complete 1–4 m.y. prior to the ignimbrite eruption and caldera collapse.

In contrast, after an initial eruption from polycyclic calderas in the San Juan region, younger ignimbrites commonly had short recurrence intervals, with precursor lavas largely confined to interiors of prior caldera basins. For example, six major ignimbrites, individually ~100–1000 km³, erupted from the Platoro caldera complex within a 1.4 m.y. span. At the central San Juan caldera cluster, seven voluminous ignimbrites erupted from sites within the large La Garita caldera during a similar interval, between 28.2 and 27.1 Ma. Three of these ignimbrites, from the San Luis caldera complex, were generated during an interval too short to resolve by ³⁹Ar/⁴⁰Ar methods, apparently less than ~40 k.y. (Lipman and McIntosh, 2008; Cantrell et al., 2023). Further documenting complex rapid magmatic evolution, diverse lava sequences (andesite-rhyolite), commonly compositionally divergent from the preceding ignimbrite, accumulated within each caldera basin between successive explosive eruptions.

Similarly in the western San Juan Mountains, four large ignimbrites erupted within only ~0.6 m.y. (29.0–28.4 Ma), although from a more diffuse cluster of calderas (Steven and Lipman, 1976; Lipman and Zimmerer, 2019). Perhaps comparably, the 0.6 m.y. interval between peak growth of the precursor Rawley volcanic complex and subsequent ignimbrite eruption at Bonanza caldera (Fig. 16B) was briefer than the inferred multi-million-year durations before the initial ignimbrite eruption from the Platoro or La Garita calderas. This contrast may be related to the spatial overlap of Bonanza with the preceding Marshall caldera (Lipman et al., 2015), together perhaps best considered as a partially polycyclic system.

The prolonged durations between peak growth of precursor lava edifices and subsequent ignimbrite eruption from a common areal locus indicate that the initial generation of a focused upper-crustal magmatic system in the San Juan region, capable of erupting a large volume ignimbrite, commonly required an extended maturation period, as much as a million years or more, during which effusive eruptions diminished. In contrast, once such a focused system became thermally mature and compositionally evolved, large volumes of silicic magma could accumulate and erupt much more rapidly.

Well-documented counterparts to the tempo of precursor eruptions at the San Juan locus are sparse. One clear analogy, though younger, areally smaller, and briefer in duration, is growth of the composite Mount Mazama edifice, which was constructed in discrete episodes prior to the 7.7 ka ignimbrite eruption and caldera formation at Crater Lake, Oregon (Bacon and Lanphere, 2006). Growth of Mount Mazama commenced at least as early as 420 ka. By 80 ka, almost the entire circumference of the edifice had grown to near-present-day levels that are preserved on the caldera rim. Dacite comparable in composition to the culminating ignimbrite was initially erupted ca. 40 ka, and although the overall growth of Mazama was nearly complete, proximal lavas continued to erupt until the culminating explosive eruption. Thus, the maturation duration for the ignimbrite magma body at Mazama was much briefer than that for San Juan systems, as were the eruptive volume, caldera area, and presumed size of the subsurface magma body. Another well-documented young caldera system is Santorini, Greece, where the lava edifice above sea level was largely constructed by ca. 450 ka, 200 k.y. before the first sizable explosive eruptions (Druitt et al., 2019).

A larger ignimbrite center, also characterized by long-lived precursor volcanism, is the Jemez Mountains volcanic field, along the west flank of the Rio Grande rift in northern New Mexico. Initial lava eruptions were under way by 13.5 Ma, a composite edifice was sizable by 8–5 Ma (Bearhead Rhyolite), and the edifice preserved on flanks of the Valles caldera had largely formed by 3 Ma (Rowe et al., 2007; Goff et al., 2011; Kelley et al., 2013). After ~1.5 m.y. of diminished activity, two caldera-forming eruptions of the Bandelier Tuff occurred relatively geologically rapidly at 1.61 Ma and 1.23 Ma (Spell et al., 1996; Nasholds and Zimmerer, 2022). Similarly at another large system, the 0.76 Ma Long Valley caldera in California, surface basalt-dacite volcanism peaked between 3.9 and 2.6 Ma and then nearly ceased during the 1.8 m.y. prior to ignimbrite eruption (Bailey, 2004; Hildreth et al., 2023).

Additional sites that provide records of magma chamber maturation succeeded by runaway growth leading to sizable ignimbrite eruptions are discussed by Townsend et al. (2019) and Bouvet de Maisonneuve et al. (2021). At most of those sites, however, the documented precaldera volcanism was less areally widespread and voluminous, some eruption ages are less constrained, and apparent precursor durations were briefer than those for the San Juan region. Several geologically young examples of rapid polycyclic recurrence of large ignimbrites after initial establishment of an upper-crustal magmatic system, broadly comparable to late-erupted ignimbrites from the central San caldera cluster, are well documented. These include the Oruanui and Taupo ignimbrites (25–1.8 ka) and smaller eruptions from the Taupo center, New Zealand (Wilson, 2001; Wilson et al., 2021), multiple ignimbrites (172–3.6 ka) from Santorini in Greece (Druitt et al., 2019), and the four large eruptions (266–86 ka) from Aso caldera, Japan (Kaneko et al., 2007; Keller et al., 2023).

Volcanic dike orientations have long been recognized as potential records of regional stress geometry (Odé, 1957; Fiske and Jackson, 1972; Nakamura, 1977; Aldrich et al., 1986; Acocella and Neri, 2009). The diverse dike configurations among the eastern Conejos edifices summarized here, all undergoing peak growth during a relatively brief geologic interval at ca. 34–32 Ma early in the development of the SRMVF, suggest that these volcanoes formed while the stress field was variable and unstable during a prolonged transition from regional compressive to extensional strain.

Dikes at many of these volcanoes can be categorized into three general configurations: symmetrically radial, sector radial, and preferential alignment (Fig. 20; Table 1). Summer Coon volcano exemplifies a radial symmetry without preferential directions for a broad range of dike compositions from basaltic andesite to rhyolite. In contrast, Tracy, Jacks Creek, and the long-active dike system that projects westward from the Platoro caldera complex are confined to a geographic sector of partial radial geometry. The BLVC and Baughman volcano both display concentrations of aligned dike directions but in contrasting directions, NNW for Baughman volcano but NNE at BLVC (Figs. 3 and 9). At both these centers, andesite dikes associated with early lavas tend to have weak incomplete radial patterns, while the more silicic dikes (dacite, rhyolite) define more linear trends. The Rawley volcanic complex contains only a few mapped dikes, perhaps in part due to limited exposures on steep debris- and vegetation-mantled slopes (Lipman, 2020); they are traceable only for short distances and fail to define systematic geometric trends. The Grays Creek and Sargents centers also lack exposed dikes. Farther west, the Needle Creek, Sawtooth, and Sky City centers have preferred north or northeast alignments of accompanying dikes. Taken together, the early eastern San Juan volcanoes generated geometrically diverse dominant dike trends during the interval 34.5–32.5 Ma, concurrent with caldera-forming ignimbrite eruptions along the southern Sawatch trend (Thorn Ranch, Bonanza, North Pass), but several million years prior to inception of the ignimbrite flare-up at the San Juan locus.

A comparable mix of diverse dike geometries is displayed by several volcanic centers in the western San Juan region (Fig. 20; Table 1). The Larsen center northeast of Lake City contains a concentration of NNE-trending dikes that extend outward from the proximal breccia cone and central intrusion (Lipman, 1976). Dikes associated with the Cimarron center farther northwest have preferential WNW trends (Lipman, 1976; Johnston, 1978), while those associated with the small Matterhorn center trend northerly, and those at the western Cow Creek center are more radial (Luedke, 1972; Steven and Hail, 1989). Ages and chemical data are lacking or sparse for the western centers, however, and some, such as those defined by monzonite stocks at Carson Camp and Castle Creek (Lipman, 1976), are younger than the northeastern Conejos edifices (Sliwinski et al., 2022). The relation of these intermediate-composition intrusions, and others farther east that lack associated dikes (Blue Creek, Grays Creek; Table 1), to volcanic constructs is poorly constrained by age determinations.

The variable dike configurations at Conejos volcanoes of the SRMVF may provide insights about concurrent changes in regional stress geometry. The generation of ignimbrite flare-ups in continental-arc volcanic fields such as the SRMVF is increasingly recognized as associated with changes in plate geometry, involving destabilization of a prior flat-slab configuration concurrent with the Cenozoic flare-ups along the Cordilleran margin of the North American plate (Coney, 1978; Lipman, 1980; Best and Christiansen, 1991; Best et al., 2016). Whether involving slab rollback or other geometric adjustments in plate boundaries, such changes also promoted a transition from a compressive to weakly extensional regional stress regime. For western North America, the ignimbrite flare-ups largely preceded development of large-scale horst-and-graben structures in the Basin and Range and Rio Grande rift extensional systems. This is documented by the spreading of outflow ignimbrites widely, without major variations in thickness across areas that later became horsts and grabens (Best and Christiansen, 1991; Lipman, 2007; Henry and John, 2013).

Nevertheless, inception of the transition to an extensional regime in the SRMVF is suggested by some structural and volcanic alignments and petrologic features associated spatially with the ignimbrite flare-ups and its volcanic precursors. These include the northerly dike trends at Baughman volcano and the BLVC, the northerly alignments of calderas of the Sawatch trend, and similar orientations of keystone faults of resurgent uplifts at calderas like Bonanza and Creede. High-silica rhyolite and granite emplaced along the Sawatch trend as early as 31–30 Ma (Zimmerer and McIntosh, 2012; Mills and Coleman, 2013) also are compositions more petrogenetically characteristic of extensional settings than continental-arc magmatism (Christiansen and Lipman, 1972). We suggest that the variable dike directions at the relatively well-dated Conejos volcanoes along the eastern flank of the San Juan Mountains may record unstable recurrent fluctuations in stress geometry early during the regional transition from arc to rift. One such alternation at ca. 33 Ma may have been captured at the BLVC and Baughman centers by the transition from the weakly radial geometry of early andesite dikes to the more linear trends displayed by later dacite and rhyolite dikes.

Additionally, the asymmetries of dike distribution at the Conejos volcanoes provide some guides about the role of basement tectonics on volcano growth, independent of regional stress geometry. The eastward sector concentration of radial dikes at Jacks Creek and Tracy volcanoes (Figs. 11 and 13) is inferred to be due to buttressing by the flank of a Proterozoic basement ridge to the west. This paleohigh constitutes the southern end of an early Cenozoic (Laramide) uplift roughly coincident with the present-day Sawatch Range. A comparable interpretation, but with opposite polarity, has been suggested for the west-directed sector of radial dikes at the Platoro magmatic locus to the south (Lipman and Zimmerer, 2019).

The early San Juan volcanoes, and other centers in the SRMVF, were less silicic but more voluminous than later ignimbrites, constituting about two thirds the total volume erupted at the San Juan magmatic locus. Andesitic compositions are dominant throughout, but dacite and rhyolite become more abundant northeastward in the San Juan region. All the early volcanic rocks have high-K calc-alkaline compositions, typical of continental-arc volcanism, but they become modestly more alkalic and enriched in trace elements such as LREE, Zr, Nb, and Th from the San Juan locus northeastward into central Colorado (Fig. 18A). Such areal compositional variations, already established at the Conejos volcanoes, may reflect inception of crustal thickening and deeper levels of primary magma generation during compression and uplift driven by low-angle subduction during the early Cenozoic, augmented by mid-Cenozoic assembly of upper-crustal magma bodies (Lipman, 2021). Thick mid-Cenozoic Rocky Mountain crust and high paleo-elevations could be consistent with otherwise-perplexing evidence for widespread rapid erosion concurrent with the early SRMVF. None of the regional ignimbrites erupted from southern parts of the Sawatch trend (Thorn Ranch, Bonanza, Saguache Creek Tuffs) is preserved on flanks of the BLVC, Baughman, or Summer Coon edifices, or elsewhere farther west in the San Juan volcanic area. This absence, in contrast to widespread outflow preservation of these ignimbrites to the east of their caldera sources, suggests that the San Juan region had already become a constructional volcanic highland, probably in conjunction with rapid syneruption erosion during early volcanism in the SRMVF (Lipman, 2021).

No detailed interpretations of petrogenetic processes are explored here, but the new data on eruptive sequences and compositions at the early northeastern volcanoes are consistent with prior inferences about the magmatic evolution at the San Juan locus. The early volcanoes are inferred to record initial incubation stages in prolonged histories of incremental magma generation, accumulation, fractionation, and solidification during construction of large subvolcanic reservoirs that intermittently became sufficiently liquid to erupt large ignimbrites (Lipman et al., 1978; Riciputi et al., 1995; Parker et al., 2005; Lipman, 2007; Lake and Farmer, 2015). The large gravity low that encompasses the San Juan ignimbrite calderas has long been interpreted as recording the presence of a composite granitoid batholith (Plouff and Pakiser, 1972; Drenth et al., 2012), representing the upper-crustal remnants of an overall translithospheric magmatic system.

Such geophysical data combined with geological constraints and comparisons with tilted plutons and magmatic-arc sections elsewhere are consistent with the assembly of vertically extensive (>20 km) intermediate to silicic batholiths (with intrusive:extrusive ratios of 10:1 or greater) beneath the major Southern Rocky Mountain volcanic loci (Lipman and Bachmann, 2015). Isotopic data require involvement of voluminous mantle-derived mafic magmas on a scale equal to or greater than that of the volcanic and plutonic rocks (Farmer et al., 2008; Lake and Farmer, 2015). Early waxing-stage intrusions (35–30 Ma) that fed intermediate-composition central volcanoes of the San Juan locus are more widespread than the geophysically defined batholith; these likely heated and processed the crust, preparatory for the ignimbrite volcanism and large-scale upper-crustal batholith growth. Such mid-Cenozoic processes of batholith assembly associated with the San Juan locus and the broader SRMVF, initiated by asthenospheric input, must have caused drastic chemical and physical reconstruction of the entire lithospheric column. Although the areal extent and volume of precursor volcanism in the San Juan locus appear to have been greater than those at most similar centers elsewhere, all large-volume silicic ignimbrite eruptions are thought to have been associated with comparable translithospheric plutonic constructs (Hildreth, 1981; Lipman and Bachmann, 2015; Best et al., 2016; Cashman et al., 2017; Hildreth et al., 2023).

The general temporal-compositional progression of continental-arc magmatism in the San Juan locus, incubated by the early intermediate-composition magmatism that led to assembly of larger bodies of more silicic magma eruptible as ignimbrites, is closely paralleled by the evolution of mid- to upper-crustal caldera-scale intrusions (Lipman, 2007). For example, the ca. 10 m.y. assembly of the Cretaceous Tuolumne intrusive complex in the Sierra Nevada of California involved early emplacement of multiple small-volume batches of compositionally heterogeneous and relatively mafic magma (analogs of Conejos lavas) before increased magma flux generated larger, more homogeneous, and more silicic phases with a caldera-scale footprint (Coleman et al., 2004; Memeti et al., 2010; Ardill et al., 2018). Many granitoid plutons, including the Tuolumne phases, are interpreted to have lost substantial volumes (>40%) of interstitial melt during crystallization (Barnes et al., 2019), melt that could have fed volcanic eruptions at the scale of the San Juan ignimbrites (Memeti et al., 2021).

Newly acquired and published map, age, and compositional data for early volcanoes in the northeastern San Juan Mountains provide insights about magmatism precursor to the large continental-arc ignimbrite flare-up in the mid-Cenozoic SRMVF. Initial volcanism migrated from central Colorado to northern New Mexico between 40 and 29 Ma as part of a more regional trend of southward progression of mid-Cenozoic magmatism in the U.S. segment of the North American Cordillera. Within the San Juan locus of the SRMVF, representing the largest preserved erosional remnant and site of most intense activity, eruptions at individual edifices appear to have begun nearly concurrently at ca. 35–34.5 Ma, with peak activity at 34–32 Ma. Vertical sections through the early volcanoes are locally as thick as 1–2 km in natural exposures and drill holes, even though upper parts of edifices have been eroded. Inferred cumulative volume of the early volcanism is about twice as large as that of the later ignimbrites. Compositions range from basaltic andesite to low-silica rhyolite, with lava volumes decreasing as silica content increased.

Although the early volcanoes are widely distributed within the SRMVF, clustered edifices that became sites for subsequent caldera formation are interpreted as recording initial magmatic focusing and growth of upper-crustal magma bodies capable of generating large-volume ignimbrites. The duration of intermediate-composition eruptions and intrusions was prolonged at some Conejos centers, commonly several million years. Overall, durations of focused magmatism at ignimbrite caldera sites were longer, e.g., as much as 9–12 m.y. at the Platoro caldera area in the southeastern San Juan Mountains. Several of the large-volume ignimbrites and related caldera subsidence events, including the enormous Fish Canyon Tuff and associated La Garita caldera, formed several million years later than peak growth of associated precursor volcanoes. Focused upper-crustal magmatic systems, capable of erupting large ignimbrites, appear to have required extended maturation intervals. In contrast, the analogous duration for smaller ignimbrite magmas, such as at Crater Lake, Oregon, was briefer than for San Juan systems, as were the eruptive volume, caldera area, and inferred size of subsurface magma bodies.

Dike configurations at the early volcanoes, which were active nearly concurrently in the SRMVF, vary from symmetrically radial to more linear trends. The contrasting dike geometries are inferred to record fluctuations from compressive to weakly extensional regional stress fields, concurrent with destabilization of the prior flat-slab plate configuration that triggered ignimbrite flare-ups along the American Cordilleran margin.

Andesitic compositions are dominant in early volcanism at the San Juan magmatic locus, but dacite and rhyolite increase northeastward in the San Juan region. The early volcanic rocks have high-K calc-alkaline compositions typical of continental-arc volcanism, but they become modestly more alkalic and enriched in trace elements such as LREE, Zr, Nb, and Th from the San Juan locus northeastward into central Colorado. Such variations may reflect synmagmatic crustal thickening and deeper levels of primary magma generation, concurrent with assembly of focused upper-crustal magma bodies that could support large ignimbrite eruptions. Substantial uncertainties remain for growth histories of the precursor volcanoes, because of unexposed lower parts of edifices, eroded upper parts, limited availability of mineral phases that can be dated reliably, and lack of detailed studies, especially for western San Juan centers.

Broadly similar precursor effusive volcanism characterizes other major loci of continental-arc ignimbrite magmatism along the western American Cordilleras. Well-documented examples include the Altiplano-Puna volcanic complex in the South American Andes, the Sierra Madre Occidental region of Mexico, and the Basin and Range and Mogollon-Datil loci of Cenozoic ignimbrites in the western United States. None of these sites, however, records early volcanism as voluminous, areally widespread, well exposed, compositionally diverse, or widely dated by isotopic methods as those in the San Juan locus in Colorado or the overall SRMVF. Perhaps the data and inferences presented here, ideally along with further field and laboratory studies in the SRMVF and elsewhere, can provide improved understanding of magmatic precursors to large ignimbrite flare-ups.

Starting shortly after the advent of quantitative dating of minerals from igneous rocks by measuring the radioactive decay ⁴⁰K to ⁴⁰Ar in the midtwentieth century (Wasserburg, 1954; Evernden et al., 1957), the San Juan locus of the mid- to late Cenozoic Southern Rocky Mountain volcanic field (SRMVF) became a recurrent focus for detailed isotopic age determinations. Such geochronologic studies have been critical for unraveling the complex eruptive and magmatic history of the SRMVF, but multiple revisions of preferred isotopic decay constants as well as progressive improvement in analytical precision have led to inconsistencies for comparisons of previously published age data. Here, we compiled and evaluated available K-Ar, ⁴⁰Ar/³⁹Ar, and U-Pb ages for the early volcanic and intrusive rocks, adjusted to currently preferred calibrations (Table S1).

Prior geochronologic work is summarized briefly, as the basis for understanding the interpretive reliability and complexities for some results compiled in Table S1. Initial determinations for six eruptive units of the San Juan volcanic locus (Steven et al., 1967) were obtained soon after establishment of a K/Ar laboratory at the U.S. Geological Survey (USGS) offices in Denver, Colorado. Despite sizable analytical uncertainties (~3%), these initial ages were major innovations in documenting that the main magmatism at this vast continental volcanic area was Oligocene, rather than Miocene–Pliocene as had been interpreted previously (Larsen and Cross, 1956). Additionally, determinations on different minerals (sanidine, biotite, hornblende, plagioclase) from several individual samples, all within analytical uncertainty, provided an early confirmation of the reliability of the K/Ar method. As one outcome, samples of the Fish Canyon Tuff (largest ignimbrite of the SRMVF) from a site west of South Fork, Colorado, became a widely used interlaboratory monitor standard (Cebula et al., 1986; Kunk et al., 1985; Hurford and Hammerschmidt, 1985; Renne et al., 1998; Lanphere and Baadsgaard, 2001). The multiple crystal phases (sanidine, biotite, hornblende, plagioclase, zircon, apatite, sphene) in this crystal-rich tuff were well suited for comparative measurement by the K/Ar, ⁴⁰Ar/³⁹Ar, U-Pb, and fission-track methods.

During the past 50-plus years, ~850 volcanic and intrusive samples from the SRMVF have been dated, while analytical precision and interpretive reliability have also improved notably. The resulting record of eruptive history, perhaps unparalleled for large-volume volcanism at a continental-margin volcanic arc, has provided a critical framework for evaluation of geologic map unit correlations, eruptive tempo, petrogenetic magmatic evolution, and formation of economically important mineral deposits.

Early determinations by K/Ar and fission-track methods have now been largely superseded by ⁴⁰Ar/³⁹Ar and zircon U-Pb analyses. During this >50 yr span, isotopic decay constants, values for radiation-flux monitors, and calibrations among analytical methods have also been revised multiple times. For example, results from different laboratories and methods have been used to propose varied preferred eruptive ages for the Fish Canyon Tuff, with its diverse phenocryst phases that are widely used as geochronologic monitors. Preferred ages from sanidine phenocrysts have varied from 27.8 Ma (Lipman et al., 1970) to 28.5 Ma (Hon and Mehnert, 1983), 27.84 Ma (Kunk et al., 1985; Cebula et al., 1986; Samson and Alexander, 1987; Deino and Potts, 1990), 28.01 Ma (Renne et al., 1998; Dazé et al., 2003), 27.57 Ma (Lanphere and Baadsgaard, 2001), and 28.201 Ma (Kuiper et al., 2008) as used in this report, to 28.06 (Phillips and Matchan, 2013) and most recently 28.175 Ma (Phillips et al., 2022).

As a result, comparison of results from data published at different times requires recalculation and adjustment of dates to updated values of decay constants and standards (for ⁴⁰Ar/³⁹Ar ages), and determination of preferred ages for individual volcanic units requires assessment of the validity of inconsistent dates. Accordingly, available ages that are deemed reliable are tabulated in this summary, recalculated as needed to Fish Canyon Tuff at 28.201 Ma (Table S1). Excluded from the tabulation are: (1) K/Ar determinations for units that were later dated by higher-precision ⁴⁰Ar/³⁹Ar or U-Pb methods and (2) some early ⁴⁰Ar/³⁹Ar dates that have proved inconsistent with later-determined ages from the same unit. A few tabulated dates that are deemed unreliable because of discrepancies with other determinations from the unit or inconsistency with stratigraphy are provided in red font on Table S1.

Up through the mid-1980s, most SRMVF ages were determined by K/Ar methods on sanidine, biotite, and hornblende phases at the USGS laboratory in Denver (e.g., Steven et al., 1967; Lipman et al., 1970, 1976, 1986; Mehnert et al., 1973a, 1973b). Although analytical uncertainties were large, the recalibrated analyses generally are consistent with ages subsequently determined by higher-precision methods. Accordingly, some K/Ar dates are tabulated here for units where more recent age data are limited or unavailable. Early-determined K/Ar ages from the San Juan region were later recalculated using the 1977 International Union of Geological Sciences (IUGS) recommended decay constants (Hon and Mehnert, 1983), although the relatively large uncertainties for the early K/Ar dates reduce the need and significance for such corrections. Fission-track dates (zircon, sphene, apatite) were also reported in several of the above-cited papers but are not tabulated here because of the large uncertainties.

Starting in the mid-1980s, in conjunction with initiation of the Creede Scientific Drilling Project (Bethke and Lipman, 1987; Bethke and Hay, 2000), an intense effort was initiated at the USGS geochronology laboratory in Menlo Park, California, to obtain higher-precision eruption ages by recently developed ⁴⁰Ar/³⁹Ar methods (Dalrymple and Lanphere, 1971, 1974), focusing on volcanic rocks of the central San Juan caldera complex (Lanphere, 1988, 1996, 2000; Lipman, 2000, 2006). Many analytically reproducible ages were obtained for the major ignimbrites and associated lavas, using total fusion, incremental-heating plateau, and isochron methods on small-volume multigrain concentrates of crystal separates (typically a few hundred milligrams or less). This may have been the most detailed and comprehensive dating effort for mid-Cenozoic volcanic rocks attempted anywhere up to that time: 114 phenocryst dates were determined by several ⁴⁰Ar/³⁹Ar (incremental heating, total fusion, laser fusion) and K/Ar methods from 59 samples. Multiple samples from different sites were dated for several major ignimbrites, and several phenocryst phases were analyzed from some individual samples. Unfortunately, the results proved to be difficult to interpret. In contrast to coherence among the field stratigraphic interpretations, petrologic data, and paleomagnetic results, the extensive USGS geochronologic results were puzzlingly inconsistent in relation to analytical uncertainties of the ⁴⁰Ar/³⁹Ar methods (then commonly ~0.5% or ±0.14 m.y. for a 28 Ma age), especially for rocks of the San Luis caldera complex (Lipman, 2000, 2006). Discrepancies substantially greater than reported analytical uncertainty, as much as 1 m.y., emerged among geographically diverse sample sites for individual outflow ignimbrite sheets, between intracaldera and outflow portions of the same ignimbrite, for multiple phenocryst phases from individual samples, and among determinations for some phenocryst samples analyzed both by incremental-heating and laser-fusion methods (fig. 11 in Lipman, 2000). The ⁴⁰Ar/³⁹Ar calibration age inferred for eruption of the Fish Canyon Tuff, 27.57 ± 0.18 Ma (Lanphere and Baadsgaard, 2001), also was substantially younger than that proposed by other isotope laboratories. Particularly puzzling in relation to geologic relations, some age data, especially for sanidine phenocrysts, appeared to suggest that eruptions from the San Luis complex were younger than those from the Creede caldera (Lanphere, 1988, 2000).

Concurrent with these USGS ⁴⁰Ar/³⁹Ar determinations in the 1990s, sanidine ages from a suite of representative samples from the Platoro caldera complex were determined by Alan Deino at the Berkeley Geochronology Laboratory, using the newly developed single-crystal laser-fusion method (SCLF). In contrast to the problematic USGS ⁴⁰Ar/³⁹Ar results on mineral concentrates from central San Juan rocks, the SCLF sanidine ages for successive ignimbrites, lavas, and intrusions at Platoro were consistent with stratigraphy, analytical uncertainties were lower, and multiple ages from single ignimbrites agreed within analytical uncertainties (Lipman et al., 1996).

In part to explore the causes of the inconsistencies among the prior USGS data from the central San Juan region, a second generation of ⁴⁰Ar/³⁹Ar studies focusing on the ignimbrites that had previously yielded such puzzling results was initiated in 2004, this time by single-crystal laser-fusion techniques in collaboration with William McIntosh at the New Mexico Geochronology Research Laboratory (NMGRL; Socorro, New Mexico) using the single-detector MAP 215-50 mass spectrometer. Resulting SCLF sanidine ages from the NMGRL were completely consistent among multiple samples from individual ignimbrites and with the stratigraphically constrained sequence of units successively erupted in the central San Juan region (Lipman and McIntosh, 2008), resolving inconsistencies among prior ⁴⁰Ar/³⁹Ar data. Reasons for discrepancies among the anomalously young ⁴⁰Ar/³⁹Ar dates determined earlier at the USGS laboratory have never been resolved in detail, but the laser fusion of multiple sanidine grains, some perhaps containing degassed adhering or included glass, may have yielded the inconsistent ages. In contrast, contamination by incompletely degassed older sanidine crystals would yield discordantly older age spectra, as identified in a few SRMVF ignimbrites (McIntosh and Chapin, 2004). Accordingly, the present compilation lists only ages determined at the New Mexico laboratory for those units where the 1980–1990s USGS dates are internally inconsistent, are in conflict with geologic relations, or have large analytical uncertainties.

Highly productive collaborative studies with McIntosh and later with others at NMGRL have continued to the present (2023), yielding the largest component of the overall geochronologic data set for the San Juan magmatic locus. When analyses obtained with high-sensitivity multicollector mass spectrometers (Argus VI and Helix MC Plus) became available from the NMGRL ca. 2012, analytical uncertainty for good-quality samples, especially SCLF analyses of sanidine, further decreased to ~0.05% (commonly ~0.015 m.y. for a 30 Ma sample). At this point, differences in age determinations beyond estimated uncertainties among multiple samples from some individual volcanic units appear as likely to result from geologic processes as from issues of analytical precision. Minute excess ⁴⁰Ar in small melt or mineral inclusions, inclusion of completely degassed antecrystic or xenocrystic grains, or trace alteration in mid-Cenozoic material may be involved. Measurable variation of neutron flux among individual crystals within pits of an irradiation tray likely plays a role as well. Whatever uncertainties continue as problems from geologic uncertainties, calibration of decay constants, and interlaboratory comparisons, the large number of age determination obtained by consistent techniques from a single analytical facility has provided robust control on the timing of eruptive sequences and repose intervals between eruptions in the San Juan region.

The resulting ages from the New Mexico laboratory have provided a critical geochronologic framework for several geologic maps, interpretive papers, and field guides (Lipman and McIntosh, 2008; Lipman, 2012, 2020; Lipman et al., 2013, 2015, 2022a, 2022b; Lipman and Zimmerer, 2019, 2022; this report Tables S1 and S3). In addition to the ⁴⁰Ar/³⁹Ar results, zircon U-Pb dates for early intrusive and volcanic rocks in the San Juan region increasingly have been obtained by secondary ion mass spectrometry (SIMS), thermal ionization mass spectrometry (TIMS), and laser-ablation–inductively coupled plasma–mass spectrometry (LA-ICP-MS) methods. Zircon dates for intrusions lacking phases reliably determinable by ⁴⁰Ar/³⁹Ar methods (Tables S1 and S4) have been obtained at the University of Geneva (Bachmann et al., 2007; Wotzlaw et al., 2013; Curry et al., 2021), the University of Arizona (Gonzales, 2015; Gonzales et al., 2021), the Stanford–U.S. Geological Survey Sensitive High Resolution Ion Microprobe–Reverse Geometry (SHRIMP-RG) laboratory (Tomek et al., 2019; Gilmer et al., 2021), and at ETH Zurich in Switzerland (Sliwinski et al., 2022).

Samples for age and compositional determinations were collected from the most pristine available sites, mainly at rapidly cooled margins of lavas and intrusions, and they were evaluated petrographically for absence of alteration. All new ⁴⁰Ar/³⁹Ar determinations were made at the NMGRL. Prior to 2012, isotopic ratios were determined using the MAP 215-50 mass spectrometer following methods similar to those summarized by Lipman and McIntosh (2008). Ages determined after 2012 were generated with the Argus VI and Helix MC Plus multicollector mass spectrometers using similar methods to those in Lipman and Zimmerer (2019, their table S5). Most previously unpublished ⁴⁰Ar/³⁹Ar data for Summer Coon volcano, collected with the MAP 215-50, were reduced using Mass Spec software, whereas new ⁴⁰Ar/³⁹Ar data for the Biedell-Lime and Baughman volcanoes were collected and reduced using PyChron (Ross, 2019). Analytical parameters, age calculations, and probability plots are listed in Table S3.

All previously published and newly listed ⁴⁰Ar/³⁹Ar ages for the pre-ignimbrite volcanism (Table S1) have been revised by simple ratio recalibration to a standard Fish Canyon Tuff age (sanidine) of 28.201 Ma (Kuiper et al., 2008). Although less accurate than full recalculation based on the decay constants and neutron flux parameters as originally reported, such data are not available for all determinations, and the discrepancy between the methods is significantly less than the 2σ uncertainties reported for mid-Cenozoic rocks. Ranges of values beyond analytical uncertainty among multiple samples from some individual volcanic units, even when analyzed at the same laboratory, indicate that geological variability can generate larger complexities than alternatives for recalibration. Originally reported and revised ages are listed with analytical uncertainties at two sigma (2σ; 95% confidence interval). The relative reliability of the individual compiled ages has also been evaluated qualitatively, based on consistency with other determinations for the same unit, accord with stratigraphic sequence, and size of analytical uncertainty. Ages deemed most reliable have been highlighted in bold font on Table S1; suspect values are listed in red font. Pooled ages for multiple reliable samples from a single unit are listed as weighted means, with uncertainties as the standard error of the mean (S_e).

New U-Pb zircon dates were determined by SIMS, using the Stanford–USGS SHRIMP-RG with methods similar to those described in Gilmer et al. (2021). Zircon grains were separated by conventional methods, mounted in epoxy, and polished to reveal interiors. Cathodoluminescenc e (CL) images were acquired with the JEOL 5800 LV scanning electron microscope at the USGS Denver Microbeam Laboratory to identify internal structure, inclusions, and physical defects. CL zoning was predominantly oscillatory with some sector zoning present as well. Zircon fragments were the most common morphology; however, some prismatic grains were present in each of the samples.

Analytical spots ~25 μm in diameter were sputtered using an ~3.5 nA O₂^– primary ion beam. Positive secondary ion intensities were measured by a single electron multiplier collector. The following peaks were measured sequentially: ⁸⁹Y, ¹³⁹La⁺, ¹⁴⁰Ce⁺, ¹⁴⁶Nd⁺, ¹⁴⁷ Sm⁺, ¹⁵³Eu⁺, ¹⁵⁵Gd⁺, ¹⁶³Dy¹⁶O⁺, ¹⁶⁶Er¹⁶O⁺, ¹⁷² Yb¹⁶O⁺, ⁹⁰Zr₂¹⁶O⁺, ¹⁸⁰Hf¹⁶O⁺, ²⁰⁴Pb⁺, a background measured at 0.045 mass units above the ²⁰⁴Pb⁺ peak, ²⁰⁶Pb⁺, ²⁰⁷Pb⁺, ²⁰⁸Pb⁺, ²³²Th⁺, ²³⁸U⁺, ²³²Th¹⁶O⁺, ²³⁸U¹⁶O⁺, and ²³⁸U¹⁶O₂⁺. The raw data were processed using SQUID 2.51 (Ludwig, 2009) and corrected for dead time. SQUID rejects data outside of the 95% confidence interval for a single population using the mean square of weighted deviates (MSWD) statistic. Measured U/Pb ratios were calibrated against the TEMORA-2 zircon standard (²⁰⁶Pb/²³⁸U age of 416.8 ± 1.3 Ma; Black et al., 2004) using Pb/U-OU/O (Williams, 1997). Concentrations of U and Th were normalized against the MAD-559 zircon standard (Coble et al., 2018) based on the measured intensity of ²³⁸U⁺/⁹⁰Zr₂¹⁶O⁺. Ratios were derived from weighted averages of within-spot scans. Individual dates and weighted means were calculated using Isoplot 3.76 software (Ludwig, 2012). Measured ²⁰⁶Pb/²³⁸U was corrected for common Pb using ²⁰⁷Pb, assuming ²⁰⁶Pb/²³⁸U-²⁰⁷Pb/²³⁵U concordance, and ²⁰⁷Pb/²⁰⁶Pb was corrected using ²⁰⁴Pb with ²⁰⁷Pb/²⁰⁶Pb values derived from the Stacey and Kramers (1975) evolution model. All ²⁰⁶Pb/²³⁸U zircon dates are reported in millions of years before present (Ma); 2σ uncertainties are noted as the total uncertainty, including the decay constant and standard (Jaffey et al., 1971) suitable for intermethod comparison of dates. Trace-element concentrations were determined using the MAD-559 standard and the reported mass fractions in Coble et al. (2018). Quantitative data for U-Pb zircon analyses and new whole-rock geochemistry are also available via USGS data release (Gilmer et al., 2023).

Locations of samples were taken from cited references wherever available. Those collected since ca. 2005 typically were recorded by global positioning system (GPS) methods. Locations have been variably listed in decimal-degree, degree-decimal-minute, degree-minute-seconds, or Universal Transverse Mercator (UTM) coordinates, most referenced to the North American Datum of 1927 (NAD27), but some to the North American Datum of 1983 (NAD83). Compiled sample locations have been adjusted to degree-decimal-minute coordinates in Table S1. While preparing data tables, a few obviously incorrect locations, presumably because of recording errors during sampling or initial compilation, were identified and corrected; other erroneous coordinates may remain unrecognized.

Locations of earlier-collected samples are less precise. Those collected by Lipman were located as precisely as possible on USGS 7.5 min topographic quadrangle maps and then converted to degree-decimal-minute coordinates by use of a Mylar® overlay with 0.1 min grid intervals. Some samples collected in the 1960s–1970s by others were only recorded by reference to distance from topographic features or other landmarks (during those decades, USGS database submittal only required listing the SE corner of the relevant topographic quadrangle map). Locations of some early-collected samples have been revised to provide more precision where possible from field notes and personal observations; others remain more approximate.

Determining reliable ⁴⁰Ar/³⁹Ar ages has been challenging for the early volcanoes because many of the intermediate-composition rocks lack phenocryst minerals for accurate dating. Ages from sanidine (single-crystal laser-fusion method) have proved to be especially reliable for determining volcanic evolution in the SRMVF, but this phase is infrequently present at the early San Juan volcanoes. Ages for most sanidine crystals from ignimbrites typically cluster within analytical uncertainty; sparse outliers are typically less than 1% older or younger than the weighted mean age. Rare older crystals from bulk ignimbrite samples commonly correspond to ages of underlying volcanic strata and accordingly are interpreted as accidental xenocrysts incorporated in pyroclastic flows during eruption and emplacement. Such outlier xenocrystic ages are typically rare to absent in sanidines from nonfragmental samples (lavas, dikes). Notably, weighted mean sanidine ages for multiple samples from single eruptive units invariably agree within or close to analytical uncertainty, and no inconsistencies with mapped stratigraphic sequence have emerged among the ~250 sanidine ages for rocks of SRMVF analyzed to date at the NMGRL.

Because of the lack of sanidine, many determinations for the early volcanoes necessarily are multicrystal step-heating plateau or isochron ages from biotite and hornblende phenocrysts, or from groundmass concentrates from samples 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 2σ (Fleck et al., 1977). Step-heating analyses for some biotite, hornblende, and groundmass concentrates yielded discordant spectra interpreted to be one or a combination of factors, including minor alteration of interstitial glass, ³⁹Ar recoil related to neutron radiation, or, for some samples, excess radiogenic Ar. Inverse isochron mineral ages were calculated for some samples that contained excess ⁴⁰Ar (⁴⁰Ar/³⁶Ar > 295.5; Nier, 1950). In cases where step-heating yielded a discordant spectrum but an acceptable array of data on the isochron with an atmospheric intercept, the isochron age was used. Most of the isochron ages agree, within 2σ analytical uncertainty, with plateau or integrated ages for the same sample, but some are younger. Results are reported as integrated (or total gas) ages for a few samples where the spectrum failed to meet the criteria for a plateau age, and the inverse isochron did not display a coherent array of data.

Many biotite and hornblende dates tend to be older than sanidine or other ages from the same eruptive unit, as indicated by inconsistencies with stratigraphic or intrusion sequence and discordance among different mineral phases from the same sample. Where multiple samples from a single eruptive unit have been dated, sanidine ages cluster tightly, while biotite and hornblende are more variable. Dates determined from some biotite or hornblende phenocrysts tend to be anomalously old for the few samples where sanidine could also be dated, and some biotite-hornblende pairs from the same sample yielded inconsistent dates (Table S5), perhaps due to varied retention of radiogenic Ar derived from Proterozoic basement rocks (Hora et al., 2010).

An exceptionally clear example of discordance is provided by high-precision Argus VI ages from the glassy basal flow breccia of a late-erupted sanidine-bearing dacite of Baughman volcano (Table S5, sample 19L-13). Repeat biotite analyses differ by nearly 1 m.y. and are 1–2 m.y. older than the preferred sanidine age of 32.13 ± 0.035 Ma for this sample. Repeat hornblende analyses are also older than the sanidine, by as much as 0.65 m.y. Vitrophyres, sampled to obtain least-altered mafic phenocryst phases as most suitable for analysis, seem to be especially prone to yield anomalously old biotite and hornblende dates relative to sanidine. In contrast, samples from the interiors of thick lavas and the few similarly dated intrusions, which presumably cooled more slowly, tend to have hornblende and biotite ages more consistent with sanidine ages. We infer the preserved presence of heterogeneous excess radiogenic Ar derived from Proterozoic basement in the vitrophyres, although such a large difference among mineral ages with relatively high analytical precision from a single sample remains perplexing.

Many groundmass determinations from andesite that lacked datable phenocryst phases have geologically plausible ages, many consistent with loose stratigraphic constraints. Some lavas yielded anomalously young dates that are inconsistent with stratigraphic succession, presumably due to Ar loss from poorly crystallized or otherwise altered interstitial sites where potassium was concentrated. Anomalous results are particularly evident for a few andesite lavas at Jacks Creek volcano, which yielded groundmass dates as much as 6 m.y. younger than mineral ages from crosscutting dikes (Fig. 14). Similarly, groundmass dates determined from some andesite breccias at Summer Coon volcano are as much as 5 m.y. younger than overlying dacite lavas and intruding dikes (Fig. 10). Nevertheless, for the relatively few samples (7) from which both mineral and groundmass were dated, most age pairs agree within or close to analytical uncertainties (Table S6). Likewise, 9 of 12 groundmass ages for the much-sampled early andesites at Summer Coon volcano and all four from the adjacent Baughman volcano are grouped within the interval 33–32 Ma (Tables S1B and S1C).

In addition to the ⁴⁰Ar/³⁹Ar results, recently determined zircon U-Pb dates for some early intrusive and volcanic rocks in the San Juan region, obtained by SIMS methods at the Stanford–USGS SHRIMP-RG laboratory, have provided valuable additional age control for samples lacking phases reliably datable by ⁴⁰Ar/³⁹Ar methods (Tables S1 and S4). Several additional U-Pb dates for SRMVF lavas and intrusions were provided prior to publication by Jakub Sliwinski; these results were determined by LA-ICP-MS at ETH Zurich (Sliwinski et al., 2022), using methods similar to those described by Guillong et al. (2014).

Additional higher-precision U-Pb dates, measured by TIMS methods for several ignimbrites of the central San Juan caldera complex (Bachmann et al., 2007; Wotzlaw et al., 2013; Curry et al., 2021), have been particularly informative. The zircon dates document prolonged histories of pre-eruption crystallization (antecrysts): from approximation of the ignimbrite eruption age as constrained by ⁴⁰Ar/³⁹Ar ages of sanidine, back several hundred thousand to more than a million years older. Analogous results have been obtained from large-volume continental-arc ignimbrites elsewhere (Reid and Vazquez, 2017; Kaiser et al., 2017; Szymanowski et al., 2019).

Comparable complexities characterize several of the eastern San Juan intrusions that yielded weighted mean U-Pb ages older by a million years or more than adjacent dated lavas. These include monzonite samples from the BLVC southern locus (19AG17), from Grays Creek (19AG14), and from the resurgent Turquoise Mine intrusion within Bonanza caldera (19AG13). While one possibility could be that these are older intrusions that were erosionally exposed and onlapped by younger lavas, a more plausible interpretation, supported by map relations and the ⁴⁰Ar/³⁹Ar lava ages, is that the zircon dates record protracted crystallization and that many of the grains are antecrysts rather than recording emplacement ages. Accordingly, we did not calculate a statistically defined single population for these samples but did include all mid-Cenozoic dates on individual zircons grains in the weighted mean. The older zircons from these samples are not morphologically distinct; they lack resorption textures or otherwise distinctive rims. The evidence for retention of earlier-crystallized zircons is especially clear for the southern monzonite of the BLVC (19AG17), where nearly 40% of zircons (11 of 29 grains) yielded Proterozoic ages (Table S4). As even more extreme examples of retained xenocrystic zircon, several mid-Cenozoic intrusions in the western San Juan region have been reported as containing only Proterozoic zircons (Gonzales, 2015).

Such results are consistent with prior interpretations that U-Pb zircon dates from intrusive rocks commonly record protracted crystallization histories, some containing large populations of antecrysts and xenocrysts (Coleman et al., 2004; Miller et al., 2007; Lipman and Bachmann, 2015; Memeti et al., 2010, 2021; Schmitt et al., 2023). Accordingly, the calculated weighted mean zircon dates that are commonly reported from SIMS analyses of volcano-related intrusions do not necessarily record reliable times of shallow crustal emplacement and final solidification. Alternative inferences, for example, that the youngest zircon dates from intrusions approximate emplacement age, may also be complicated by potential for lead loss from slowly cooled intrusive rocks, especially for samples that have been analyzed by SIMS methods without chemical-abrasion pretreatment.

As in other recent reports for the San Juan region (e.g., Lipman et al., 2015; Lipman and Zimmerer, 2019; Lipman, 2020), rock names are used in general accord with the IUGS classification system (Le Bas et al., 1986); in particular, the term “dacite” is used for rocks designated as latite or quartz latite in some prior publications. Volcanic rocks along the eastern San Juan flank constitute a high-K assemblage that is transitional between subalkaline and alkaline suites, similar to those throughout the SRMVF (Fig. 18A). For simplicity and continuity with previous usage, such modifiers as “high-K” or “trachy” are omitted from most rock names. Names divided on the basis of percent SiO₂ are: <52, basalt; 52–57, basaltic andesite; 57–62, andesite; 62–66, dacite; 66–70, silicic dacite (formerly quartz latite); 70–75, rhyolite; and >75, silicic rhyolite (with all compositions for bulk-rock analyses recalculated to reported summations volatile-free, and all FeO as Fe₂O₃).

Other than obvious xenocrysts, the megascopically visible crystal cargos of volcanic rocks are referenced as phenocrysts in unit descriptions, without necessarily implying equilibrium with compositions represented by the groundmass matrix. Many Conejos rocks display textural and other evidence of disequilibria among phases in the crystal cargo and between crystal and groundmass compositions. Total phenocryst content tends to increase with SiO₂, from andesite to dacite; plagioclase is the most abundant, accompanied by clinopyroxene and (or) hornblende in andesite. Biotite appears in silicic andesite and is the primary mafic phenocryst in dacites. Coarsely porphyritic plagioclase andesite (platy-plagioclase andesite) is a distinctive lava type that is transitional in composition to mafic dacite. Sanidine is present in a few silicic dacites and many rhyolites. Rhyolites tend to be crystal poor (<10%), everywhere containing plagioclase and sparse biotite ± quartz. Such descriptive distinctions, based on chemistry and phenocrysts, tend to obscure the gradational textures and compositions among the volcanic rocks. Uncertainties in hand-lens–based field determinations have likely led to local inconsistencies in map nomenclature, especially for andesitic and dacitic lavas, which were not mapped or subdivided as rigorously as the more silicic lavas and ignimbrites.

Chemical and petrographic data for the study areas include 65 new major-oxide and trace-element analyses. Bulk-rock compositions were determined by XRF and ICP-MS methods in laboratories at Washington State University (https://environment.wsu.edu/facilities/geoanalytical-lab/) and the Hamilton Analytical Laboratory, Hamilton College, New York (https://www.hamilton.edu/academics/analytical-lab). All major-oxide analyses were recalculated volatile-free to originally reported analytical totals. Despite potential alteration issues, most samples yielded plausible magmatic values, even for especially mobile elements like the alkalis. Effects of alteration are evident, however, in the modestly greater data scatter on data plots than is typical of young volcanic suites. A few obviously anomalous samples with suspect compositions (marked in red in Table S2) were omitted from the plots of chemical data.

We thank William McIntosh for initiating the challenging dating effort at Summer Coon volcano in 2006–2007. Jakub Sliwinski provided results of his U-Pb zircon age determinations before publication. Lisa Peters and numerous students at the New Mexico Geochronology Research Laboratory helped with mineral separates and data collection. Mathew Granitto guided access to unpublished chemical analyses of western San Juan volcanic and intrusive rocks, as tabulated in the U.S. Geological Survey (USGS) National Geochemical Database. Perceptive manuscript comments were provided by USGS reviewer Ren Thompson, Don Parker, and an anonymous Geosphere reviewer, and especially Geosphere Associate Editor Eric Christiansen. We also thank friends in the northeastern San Juan region who provided diverse hospitality, logistical support, help with backcountry and property access, and other assistance. These especially include Pip and Aaron Conrad of the Rafiki Ranch near Villa Grove, who created a virtual “Rafiki Ranch Geologic Research Station” and hosted many visiting earth scientists. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. government.