Controls on hydrothermal fluid flow in caldera-hosted settings: Evidence from Lake City caldera, USA

Calderas have long been recognized as hosts for hydrothermal systems that can be economically important for geothermal power and the localization of ore deposits (e.g., Smith and Bailey, 1968; Rytuba, 1976; Duex and Henry, 1981; Stelling et al., 2016). In addition, characterizing hydrothermal activity in large, modern calderas worldwide (e.g., Yellowstone, USA, Campi Flegrei, Italy, Taupo, New Zealand) has been important in understanding their restless behavior and associated hazards (Peltier et al., 2009; Rinaldi et al., 2010; Hurwitz and Lowenstern, 2014). These systems may be associated with the margin of a caldera, its interior, adjacent regional-scale structures, or some combination thereof (Table 1). It is not well understood why some calderas host hydrothermal systems while others do not or what factors promote fluid localization in certain parts of a caldera. In particular, caldera “ring faults” are commonly suggested to be important structures for localizing fluid flow (e.g., Duex and Henry, 1981; Wood, 1994; Guillou-Frottier et al., 2000; Stix et al., 2003; Kissling and Weir, 2005); yet no studies to date have focused on a thorough examination of their permeability structure.

Hydrothermal systems occur in a range of crustal settings. This paper is primarily focused on the upper ∼2 km of a silicic caldera-related setting, where topography has less effect on reservoir fluid flow than in stratovolcano settings (Henley and Ellis, 1983). This corresponds with the mineralization and discharge zones of Rowland and Simmons (2012), which are above the feed zone that extends down to the brittle-ductile transition and the base of convection. It is in this upper portion of hydrothermal systems that discrete high-flux fluid conduits are important; the formation of these conduits strongly depends on the interplay of structure and lithology (Rowland and Simmons, 2012; Vignaroli et al., 2015). The location of fluid conduits is, thus, expected to be influenced by caldera-related structures and lithology.

Ring faults accommodate the bulk of caldera collapse and comprise part of the structural margin of a caldera. In this study, we use the term “structural margin” for the zone of deformation at the caldera margin and the term “ring fault” for the portion(s) of this with the highest strain, if present. Fault rocks (e.g., breccia, gouge, cataclasite, and pseudotachylyte) will generally form in, and be indicative of, the highest strain (i.e., ring fault) portions of the caldera margin. The topographic margin of a caldera is where the intracaldera fill is juxtaposed against pre-caldera rocks or a significant scarp slope exists, but there is no evidence of faulting along this contact (Smith and Bailey, 1968; Spray, 1997; Lipman, 2000; Cole et al., 2005; Branney and Acocella, 2015). In many calderas, the contact between intracaldera fill and basement can be located, but exposure is insufficient to ascertain whether this defines the topographic margin or the structural margin. In this study, we call this potentially ambiguous contact between intracaldera and extracaldera rocks the “caldera margin discontinuity.” Ring faults differ from other types of faults by: (1) a high strain rate; large displacements (>1 km for some calderas) are accommodated in a short period of time (i.e., during and/or immediately posteruption) (Spray, 1997) and (2) frequent modification by magmatism (e.g., dike intrusion into the fault) and/or landsliding soon after their creation (Branney and Acocella, 2015). Well-exposed outcrops of ring faults are globally rare (Geyer and Martí, 2014), and work to date has focused on the volcanological aspects of unmineralized examples (e.g., Shannon, 1989; Moore and Kokelaar, 1998; Miura, 1999; Kokelaar, 2007).

Most faults, in general, can be broadly separated into a fault core that accommodates most displacement and consists of fault rocks such as gouge and cataclasite and a surrounding damage zone of highly fractured and deformed rock (Caine et al., 1996; Kim et al., 2004; Mitchell and Faulkner, 2009). Numerous studies have investigated the fault architecture, fault rocks, damage zones, and resulting permeability structure in diverse fault styles and protoliths (e.g., Wibberley and Shimamoto, 2002; Micarelli et al., 2006; Faulkner et al., 2010; Walker et al., 2012; Soden and Shipton, 2013). However, there have been no studies of the architecture and permeability of ring faults, despite the common association of hydrothermal fluid flow with caldera margins (Table 1).

Calderas are complex volcano-tectonic structures, and ring faults are not the only type of fault important in these settings. To understand controls on hydrothermal fluid flow in such settings, other factors such as lithology, intracaldera structures and regional structures must be considered as well as the architecture of the caldera margin. Resurgent uplift of the interior of a caldera often causes smaller-displacement normal faults to form on the crest of the resurgent dome (Smith and Bailey, 1968). Faults associated with regional tectonics are also common in caldera settings and accommodate caldera subsidence and/or resurgent uplift in some cases (Cole et al., 2005; Folkes et al., 2011). Hydrothermal fluid flow may be associated with any combination of ring faults, intracaldera faults, or regional faults in calderas worldwide (Table 1).

Ignimbrites are the dominant lithology in most silicic calderas, and the intracaldera facies of many ignimbrites are densely welded and have low intrinsic permeability (Heap et al., 2014). Faults or horizons of more permeable lithologies (such as less consolidated breccias) are important for facilitating intracaldera fluid circulation through voluminous, otherwise impermeable lithologies such as intracaldera ignimbrites (Hulen and Nielson, 1986; Wood et al., 2001). Exhumed, well-exposed calderas are needed to study structure and lithology in detail, especially to put the outcrop scale in the context of the wider structural setting. Recently active calderas are too poorly exposed—even those that have been drilled for resource exploration—to allow effective integration of data from different scales. In this paper, we present and discuss new observations and mapping of the well-exposed Lake City caldera in the San Juan Mountains of Colorado, USA. We characterize outcrop-scale patterns of veins and lithology along the margin and compare them with structures from the interior of the caldera that have much smaller displacement. Integration of map and outcrop-scale data allows us to build a conceptual model of fluid-flow localization from favorable combinations of lithology, fault and fracture density, and fault and fracture orientation. This model could aid in the prediction for where and how fluid flow (or mineralization) is localized at other calderas worldwide.

The ca. 22.9 Ma Lake City caldera (LCC) is the youngest caldera in the San Juan volcanic field (SJVF) in southwestern Colorado (Fig. 1), which is the largest erosional remnant of the Oligocene to Miocene Southern Rocky Mountains volcanic field (Lipman et al., 1970; Lipman and McIntosh, 2008). Lake City caldera postdates all other caldera-related eruptions in the SJVF by more than four million years, and it is nested within the older (ca. 28.2 Ma) Uncompahgre caldera (Bove et al., 2001). The basement lithology of LCC is the Precambrian Granite of Cataract Gulch, which is exposed along the western to southern caldera margin (Figs. 2A and 3A). The granite is coarse grained (2–8 mm) and mineralogically uniform where undeformed (Larsen and Cross, 1956; Larson and Taylor, 1986a). The granite is intruded by numerous Ordovician to Cambrian diabase dikes, which mostly have a west to northwest orientation (Lipman, 1976b). The rest of the LCC margin is bordered by the fill of the Uncompahgre caldera (Fig. 2A). The Uncompahgre caldera fill consists of welded intracaldera ignimbrite (Sapinero Mesa Tuff), landslide megabreccias of Uncompahgre caldera, and andesitic to dacitic lavas (Figs. 1 and 2A). The granite and Uncompahgre caldera fill are variably included in the Lake City landslide breccias as blocks and are deformed at the LCC margins. The deformation of the granite is described in detail later in this paper.

Lake City caldera was the source of the Sunshine Peak eruption, which produced a generally densely welded tuff (Figs. 2A and 3C) that is variably altered and, locally, lithic rich (Hon, 1987; Kennedy et al., 2015). Proximally, the tuff is closely associated with breccias (Figs. 2A and 3B) and megabreccias that formed from material collapsing from the over-steepened caldera walls during or following subsidence (Hon, 1987; Kennedy et al., 2015). The resulting caldera is oval shaped, within which ∼260 km³ of ignimbrite ponded up to a thickness of at least 1.5 km (Fig. 2A) (Hon, 1987).

Following caldera formation, magmatic activity continued with a series of domes and intrusions. Small “ring dike” intrusions were emplaced along the ring fault in places (Hon, 1987). These intrusions are highly altered and fractured (Fig. 3D). Dacitic lava domes and intrusions were emplaced along the caldera margin in the east of the caldera (Fig. 2A) and syenite (Fig. 3E) and monzonite resurgent intrusions in the center of the caldera (Fig. 2A) (Kennedy et al., 2012, 2015). Syenites, monzonites, and the domes are generally less affected by postemplacement alteration and deformation. The emplacement of the resurgent intrusion caused doming of the center of the caldera with a northeast-southwest–oriented apical graben and inversion of displacement on the Alpine Gulch fault and parts of the reactivated ring fault. The largest displacements were ∼1 km, near the intersection of the Alpine Gulch fault and the ring fault (Hon, 1987).

Mining in LCC has been less productive than in the Silverton, Uncompahgre, and San Juan calderas and nearby plutons (Slack, 1980). However, there is still evidence for widespread alteration and mineralization from hydrothermal fluids (Hon, 1987). Alteration and veining are generally most intense in the center of the caldera, at Red Mountain on the eastern edge of the caldera, and in discontinuous zones near the caldera margin (Larson and Taylor, 1986a, 1986b; Hon, 1987; Sanford, 1992). Oxygen-isotope data show that the basement granite is progressively more altered closer to the margin of the LCC (Fig. 2B) (Larson and Taylor, 1986a). Meteoric water is the principal source of hydrothermal fluid in shallow systems such as Lake City. Almost all meteoric water has a δ¹⁸O composition less than 0‰ (Vienna standard mean ocean water [VSMOW]); hence, the more fluid-rock interaction that occurs, the lower the resultant δ¹⁸O composition of the rock (Taylor, 1974). Thus, the δ¹⁸O composition of alkali feldspar in the granite varies from more than +7‰ away from the caldera margin to less than +1‰ close to the western and southwestern portions of the caldera margin (Fig. 2B) (Larson and Taylor, 1986a, 1986b). The other areas of lowest whole-rock δ¹⁸O (from +3‰ to a minimum of –1.8‰) are in the center of the caldera inside and near the resurgent intrusions (Fig. 2B) (Larson and Taylor, 1986b). Within the caldera, most veins and faults have a northeast orientation, which is related to the adjacent northeast-trending Eureka graben formed by resurgence of the Uncompahgre–San Juan caldera (Fig. 2A) (Hon, 1987). This northeast orientation is also similar to the orientation of the Colorado Mineral Belt, a 500-km-long belt of plutons and mining districts stretching from the Four Corners area in southwestern Colorado to an area near Boulder (Fig. 1). The Colorado Mineral Belt passes through the Lake City area and is related to Precambrian basement weaknesses that cut across the younger geologic grain of Colorado (Chapin, 2012). The Uncompahgre caldera, Colorado Mineral Belt, Eureka Graben, Lake City caldera, and its apical graben all trend northeast-southwest (Figs. 1 and 2). This may reflect underlying basement structures, some of which may have been reactivated during caldera collapse and resurgence. However, most faults of the apical graben were probably new structures formed during resurgent doming due to their short trace lengths and restriction to the Lake City caldera fill (Hon, 1987).

The timing of mineralization is complex in the SJVF, and the Lake City area is no exception. Some mineralization in the larger Uncompahgre–San Juan caldera is older than the collapse of LCC. This includes the Golden Fleece vein and early stages of the Ute–Hidden Treasure system, adjacent to the caldera on the eastern and northern sides, respectively, and probably related to resurgence of Uncompahgre caldera (Fig. 2A) (Lipman et al., 1976; Hon et al., 1985; Bove et al., 2001). Mineralization at Red Mountain and late stages of the Ute–Hidden Treasure system are close in age to the formation of LCC at 22.9 Ma (Bove et al., 2001). Some mineralization outside the southern margin, at the Black Wonder mine, is younger than the caldera at 16.5 Ma, as is some mineralization to the west in the nearby Silverton caldera (Lipman et al., 1976; Bove et al., 2001; Fig. 2A).

The veins, alteration, and mineralization at LCC leave a record of hydrothermal fluid flow inside and outside the caldera. The diversity of well-exposed structures (ring fault, intracaldera normal faults, and faults outside the caldera) and lithologies make this an ideal setting to build a conceptual model of the factors that promote fluid flow in calderas, particularly at their margins.

Lithological, structural, and vein mapping was focused on two areas of the caldera margin—known locally as “Shelf Road” in the southwest corner and “Red Gulch” on the western side (Fig. 2). Identification of fine-grained, granite-derived fault rocks was aided by thin-section petrography. Mapping revealed the caldera margin architecture and its relationship to hydrothermal veins in greater detail than the pioneering maps of Peter Lipman (1976b) and Ken Hon (1987).

Transects (or “scanlines”) were made across key structures with good exposure, following the methodology of Manda and Mabee (2010). Nine scanline surveys were made—two in the Shelf Road area on the caldera margin (SW2 and SW1), two at Red Gulch on the caldera margin (W1 and W2), and five inside the caldera. Of the scanlines inside the caldera, four (I3, I4, I5, and I1) were in Sunshine Peak Tuff, while one (I2) was in the syenite (Fig. 2). The scanlines varied in length from 5 m to 64 m, depending on the available exposure, and all veins with a trace length of >0.2 m were measured. This cutoff was chosen because surface iron-oxide staining adjacent to veins obscures the location of some veins with a trace length of less than 0.2 m. It was therefore not considered possible to accurately record the location and orientation of all veins with such a short trace length. The orientation and location of every intersecting vein were recorded. Non-mineralized fractures were not recorded, as these do not record the presence of hydrothermal fluids. Veins that have an orientation subparallel to the scanline will be intersected less often than features that have an orthogonal orientation; therefore, subparallel structures are underrepresented in scanline data. To account for this, the Terzaghi (1965) correction was applied, using the software Stereonet 9 (Allmendinger et al., 2012; Cardozo and Allmendinger, 2013). For the correction, we used the average orientation of each scanline and a “maximum correlation” (weighting) factor of 2. Stereonet 9 was also used to plot and analyze structural data, perform kamb contouring of poles to planes (Kamb, 1959), and calculate the circular variance of scanline strike data (e.g., Allmendinger et al., 2012; Cardozo and Allmendinger, 2013). Circular variance is a measure of the spread of the strikes of veins. It is a value between 0 (very well clustered) and 1 (very poorly clustered).

We describe the detailed distribution of fault rock facies (Table 2), subsidence structures, and veins in two areas where the caldera margin is well exposed. Additionally, we present four scanlines in these two areas of the margin and five additional scanlines from inside the caldera (Table 3) across faults where the entirety of the fault core is replaced or cemented by vein material. All veins that were recorded on scanlines, and most that were mapped, contain predominantly quartz with varying but always lesser amounts of calcite, adularia, and pyrite. Euhedral “comb” quartz grains, with relict “vugs” of open space in the center of the vein, are common. Most veins appear to be extensional (mode I) and syntaxial (inward growing; see Bons et al., 2012), though unequivocal kinematic indicators are rare. Alteration is generally propylitic on the western caldera margin and propylitic or phyllic in the interior and is usually more pervasive in a “halo” around veins. The halos vary in width from several millimeters to several meters.

The granite is phaneritic with mostly 2–8 mm crystals of quartz, alkali feldspar, and plagioclase with minor muscovite and biotite (Fig. 3E). Near the caldera margin, the granite is commonly deformed and brecciated (see below). The granite is altered near the southwest caldera margin and in the Eureka Graben (Fig. 2). The granite mostly alters to illite, calcite, and quartz.

Field mapping revealed previously unrecognized brecciated and cataclastic granite (Fig. 4 and Table 2) in some locations near the Lake City caldera margin (Figs. 5A and 5B). These fault rocks have been divided into four “facies” (facies A to D), reflecting progressively higher levels of brecciation and shearing (Fig. 4 and Table 2). The fault rocks are named according to the classification scheme of Woodcock and Mort (2008). Not all four facies are present at every location, and they all exhibit a degree of alteration and host hydrothermal veins—an indication of their coherent nature. Table 2 describes the fault rocks and their distribution in more detail, and they are shown in Figure 4.

In the vicinity of the Shelf Road, intracaldera rocks of the Lower Sunshine Peak Tuff and associated landslide breccias are in contact with the basement lithologies, which consist of Proterozoic granite and diabase dikes that have been interpreted to be Ordovician (Lipman, 1976b) (Fig. 5D). Some small pegmatite dikes are also present. The caldera margin had previously been mapped as simple and arcuate in this area; however, our field mapping (Fig. 5A) has revealed two largely straight segments with a distinct corner between them with overshooting fault zones and a splaying and reconnecting fault segment. A large fault and vein structure (Fig. 5C) links this corner on the caldera margin with a diabase dike in the granite basement. Veins and breccias suggest that the margin of this dike may be a continuation of the ring fault where it continues roughly subparallel to the caldera margin to the SE.

Granite breccias (facies A and B; see Table 2) are found close to the caldera margin discontinuity at most locations where this margin is sufficiently exposed. There is a general progression toward coarser clast sizes farther from the margin. Veins (predominantly quartz and calcite) are generally more common in the granite basement than in the intracaldera rocks (Fig. 5D) and generally decrease in density away from the caldera margin discontinuity. Veins crosscut all fault rock facies, and no veins were found in the granite more than ∼2 km from the caldera margin discontinuity (Fig. 5A). The orientation of veins outside the caldera is variable but with a maximum in the NE direction.

Near the Shelf Road in the southwestern corner of the caldera, the caldera margin orientation changes from approximately east-west to northwest-southeast (Fig. 5A). The veins in scanlines SW2 and SW1 have orientations well clustered around the east-west orientation of the caldera margin (Table 3). The orientation of veins measured in these scanlines (Fig. 5B) is much less variable than the orientation of surrounding mapped veins in the granite (Fig. 5A). In both scanlines, vein density is highest near the main caldera–related structure (Fig. 5B and Table 3). In scanline SW2, we interpret the intracaldera breccia in the ∼5 m adjacent to the caldera margin discontinuity as fault core. It has a higher proportion of matrix and finer clast size than the rest of the intracaldera breccia at this location. These breccias are therefore consistent with a landslide origin for intracaldera breccias that have undergone some dynamic textural modification by faulting, as described by Hon (1987) and Kennedy et al. (2016). Most veins on scanlines SW1 and SW2 have a euhedral comb texture indicating the quartz minerals grew from the vein wall inward toward the vein (syntaxial) into open space, with no evidence of shearing or offset (Fig. 6). These are interpreted as extensional (type I) veins (cf. Bons et al., 2012) and have an east-west orientation on average, though there is significant variability in vein strike.

The western caldera margin discontinuity at Red Gulch (Fig. 7A), consists of intracaldera rocks of the lower and middle Sunshine Peak Tuff and associated landslide breccias in contact with basement granite overlain by intracaldera rocks of the Uncompahgre caldera. The basement granite is intruded by small gabbro and pegmatite dikes. The caldera margin zone here has a generally north-south orientation and contains clastic rocks, veins, and a porphyritic rhyolitic “ring dike” intrusion (Fig. 7D). Most of the clastic rocks consist of monomict granitic breccias (facies B), with one area where the clasts are dominantly andesite. We interpret a thin (2-m-wide), north-south–trending zone of meso- to ultracataclasite (facies D) and discontinuous veins to represent the highest strain portion of the ring fault of the caldera (Fig. 4D). The breccias become finer grained toward this cataclasite zone and grade into protocataclasite (facies C) with a shear fabric (Figs. 4C and 7C). A short distance (20–30 m) away inside the caldera, the breccias are dominated by andesite and tuff clasts, as is common for most of the intracaldera landslide breccias (Lipman, 1976a). The contact between the granite-dominated breccias (fault rock facies B) and the landslide breccias is sharp, and it crosscuts the fabric of facies C slightly to the north (Fig. 7A). We interpret this as an erosional boundary formed by destruction of fault rocks from landsliding of the fault scarp into the caldera. Outside the caldera and near the cataclasite zone, the granite country rock is deformed (facies C) and contains a shear fabric in places. Most of the veins at Red Gulch are close to the caldera margin or outside of it, and veins crosscut all fault rock facies. Veins of variable orientation pervade the “ring dike,” while outside of the caldera, the majority of large veins have a northeast orientation. The latter observation is consistent with the location of this site within the northeast-trending Eureka Graben.

Exposures at Red Gulch on the western edge of the caldera indicate that the caldera margin is oriented approximately north-south and cuts across the Eureka Graben, in which many faults and veins are oriented northeast-southwest, as in the rest of Lake City caldera (Figs. 2 and 7A). The scanlines show that most veins within 10 m of the caldera margin have a north-south orientation, rather than following the regional northeast-southwest trend, and vein thickness is greatest near the ring dike margin (Fig. 7B and Table 3). Veins on scanlines W1 and W2 are interpreted as syntaxial, extensional quartz veins. Large (>10 mm) vugs of open space are common in the center of veins, especially in scanline W2. No shear or hybrid veins were identified.

The intracaldera mapping focused on areas of veins and hydrothermal alteration. Identification of cataclastic rocks or facies was difficult due to the smaller sizes of fault cores, intense silicification, and presence of hydrothermal vein breccias. The interior of the caldera is dominated by smaller-displacement normal faults associated with intrusion and uplift and parallel to regional trends (Fig. 8A). More limited exposure across key structures in the caldera interior allowed focus on shorter scanlines (Fig. 8B). In general, less vein calcite was identified in veins in the interior of the caldera than at the southwestern margin. However, bladed calcite replaced by quartz is quite widespread.

Five scanlines were made were made across mineralized northeast-southwest–oriented faults inside the caldera (Fig. 8A). Most mapped faults inside the caldera have the same orientation (Fig. 2). From bedding offset, the total displacement of some of the faults in scanlines I1 and I3 can be constrained (Table 3). There is no obvious bedding to measure displacement for the other faults; so displacement is unconstrained. However, their trace length is shorter than the faults in scanlines I1 or I3 (Fig.8A); therefore, displacement is probably less than 120 m. The circular variance and vein density patterns for the intracaldera scanlines are summarized in Table 3. In general, the major veins that the intracaldera scanlines transected are smaller in total width compared to the scanlines near the caldera margin (Figs. 5B and 6B). Veins in two of the scanlines have well-clustered northeast-southwest orientations, while the others are less clustered (Table 3).

In summary, a wider zone of deformation (up to ∼300 m) than previously mapped has been identified at the southwest corner of Lake City caldera, with at least one large inward-dipping mineralized fault extending into the granite basement outside of the caldera fill (Figs. 5A and 5B). A sheared preexisting dike, subparallel to the caldera margin, may also have accommodated some caldera-related deformation. At the western caldera margin, there is evidence for the destruction of fault rocks at moderate to high stratigraphic levels (Fig. 7A). Deformed granite fault rocks are present at both southwest and western segments of the ring fault (Fig. 4 and Table 2), and veins are associated with the caldera margin in both. At the southwestern margin, veins are more common in the granite outside the caldera than in the intracaldera breccia (Fig. 5A). Outcrop-scale scanline surveys (Table 3) in both the west and southwest indicate that the orientation of veins near the caldera margin is similar to the orientation of the caldera margin itself (Figs. 5B, 7B, and 9A). Veins surrounding smaller-displacement faults in the center of the caldera generally show orientations that are more variable and are not necessarily similar to the orientation of the faults they are associated with (Figs. 8B and 9A). The width of veins is generally larger near the caldera margin discontinuity than inside the caldera and surrounding smaller intracaldera faults (Fig. 9B). Figure 9 provides a comparative summary of scanline data from the three parts of the caldera studied. The average thickness of vein per 0.5 m (Fig. 9B) is calculated by using only the first 4 m of each scanline away from the relevant major structural discontinuity, so that the shorter intracaldera scanlines can be compared with those on the margin.

The hazard and resource potential of calderas has led to intensive research on the mechanisms of caldera formation and the geometry of the resulting structures. Because a large caldera-forming eruption has never been observed, analogue and numerical models are useful tools to investigate the variables that control the diverse caldera morphologies. For any model, it is important to validate results by comparing them with actual field examples, and detailed lithological and structural mapping of well-exposed calderas can provide valuable comparative data. The first section of this discussion focuses on the interpretation of the geology of Lake City caldera, its controls, and comparisons with analogue models and other global examples. One of the important applications of understanding caldera structure is being able to predict the resource potential of calderas. As such, the second section of this discussion focuses on the potential relationship between structure, lithology, and fluid flow at the Lake City caldera and how this compares with other calderas.

At Lake City, the caldera margin is a structurally complex zone of deformation, and we, therefore, prefer to describe it as a caldera margin discontinuity instead of a simpler ring fault, sensu stricto. We use the general term caldera margin discontinuity to describe the interface between intracaldera and extracaldera rocks where it is ambiguous if the interface is structural (i.e., ring fault) or topographic, or if it varies between these along its length.

Detailed mapping in the Shelf Road area southwest of the caldera shows that the orientation of the caldera margin changes abruptly as a corner rather than gradual curve. This suggests that the structural margin of Lake City caldera may have a more polygonal shape rather than ellipsoidal as the previously mapped outline would suggest. In modern calderas, the most obvious morphological feature is often the topographic margin of the caldera, which often has a rounded perimeter and is located outboard of the structural ring fault margin (Lipman, 2000). The topographic margin is created by shedding of material off the walls of the structural margin, the same process that has been inferred to be responsible for the widespread intracaldera breccia units in Lake City caldera and other calderas worldwide (Lipman, 1976a; Hon, 1987; Lipman, 2000). The structural margin of a modern caldera is difficult to identify because it is usually buried by the sediment that was shed during creation of the topographic margin. It may be more rectilinear or polygonal than the topographic margin, because the fault orientations accommodating the caldera collapse tend to follow preexisting planar weaknesses in the crust (Branney and Acocella, 2015). Some examples of calderas where the structural margin has been inferred to be polygonal are Rotorua, Okataina, Taupo (New Zealand), Cerro Galán (Argentina), Colli Albani (Italy), Deception Island (Antarctica), and Sakurae (Japan) (Spinks et al., 2005; Giordano et al., 2006; Komuro et al., 2006; Folkes et al., 2011; Ashwell et al., 2013; Martí et al., 2013). Caldera subsidence experiments also indicate polygonal faults may form even in the absence of structural weakness and can propagate beyond their intersecting corners (Kennedy et al., 2004; Holohan et al., 2013). In the southwest corner of Lake City caldera, a fault-vein structure splays off the caldera margin at the corner and continues southeastward until it meets a highly mineralized diabase dike (Fig. 5A). The dike contains abundant quartz veins. We interpret breccias at its northern margin as chaotic fault breccia, according to the classification scheme of Woodcock and Mort (2008). Additionally, we interpret that the NW-SE–trending fault that splays off the caldera accommodated some caldera-related displacement. The margin of the dike probably acted as a preexisting weakness at a favorable orientation to accommodate slip, as it is subparallel to the east-west portion of the caldera margin. This shows that the caldera displacement was not all focused on just one ring fault but a more distributed area of deformation (or fault zone sensu Childs et al., 2009) at least ∼300 m wide (Fig. 5A). The fault zone width of ∼300 m for a displacement of ∼1.5 km is consistent with the (extrapolated) data of Childs et al. (2009). We suggest that the subparallel orientation of pre-caldera dikes and sections of the caldera margin reveal a common regional structural control.

In the west and southwest of the caldera and closest to the discontinuity, four “facies” of deformed granite with variations in clastic textures can be identified (Fig. 4 and Table 2). These clastic rocks show a decrease in grain size from facies A to D, with facies C and D exhibiting a clear fabric. We suggest that the decrease in grain size and formation of a fabric is due to increasing amounts of strain and deformation. These rocks are found only near the margin of Lake City caldera, and we interpret that these fault rocks formed by movement along the caldera structural margin. Not every facies is present at every location, due to variable fault core development and due to the destruction of fault cores by gravitational collapse of the margin during caldera formation. There is evidence for the latter at the western caldera margin, where well-developed foliated granitic protocataclasite is abruptly crosscut by polymict andesite and tuff-dominated breccia that we interpret was formed by landsliding (Fig. 7A). There are also large portions of the caldera margin where intracaldera landslide breccia is juxtaposed against granite country rock without any fault rocks in between; presumably the fault rocks have been destroyed by gravitational collapse. At Red Gulch, there is a small area of andesite and Sapinero Mesa tuff (from Uncompahgre caldera)–dominated breccia among the granite-derived fault rocks (Fig. 7A). This breccia does not have any kind of fabric and was intruded by the ring dike on its southern end. The origin of this andesite and tuff breccia is not clear. It is not likely to be fault derived because its composition varies markedly with the surrounding granitic breccias, and it has no fabric. Our preferred explanation is that it is the remains of an extensional “crevasse” on the margin of Lake City caldera, which formed after the formation of the granitic fault rocks and was infilled with blocks of the andesites and Sapinero Mesa tuff above the granite. Similar megabreccia-filled crevasses have been identified at Glencoe caldera (Moore and Kokelaar, 1998).

Due to the NW-SE extension direction inside the caldera indicated by numerous NE-trending normal faults, it could be expected that the northern, northwest, and southeast sections of the ring fault, which are oriented approximately NE-SW, would be favorably oriented for the creation of extensional fractures and veins. There are, however, very few veins and/or alteration along these sections, despite similar amounts of exposure as along the much more mineralized western and southwestern sections (Fig. 2; Hon, 1987). The southeast part of the caldera margin is a similar distance from the resurgent pluton as the western and southwestern portions, but in contrast, it is across strike from the predominant northeast trend. Our data indicate that fluid flow in the resurgent dome may have been enhanced in the along-strike direction (with respect to the NE-SW–trending normal faults), utilizing abundant intersecting structures, and comparatively inhibited across strike toward the southeast part of the caldera. Permeability of the northern and northwestern margins may have been inhibited (compared to the western and southwestern portions) due to the lack of fault intersections (Fig. 2). Additionally, reactivation of an outward-dipping ring fault in this section of the caldera during resurgence may inhibit permeability in a manner similar to that proposed for Colli Albani caldera in Italy (Giordano et al., 2006). However, we do not have any data on the dip direction of this portion of the ring fault; so this inference is purely speculative.

The prolonged history of mineralization in the vicinity of Lake City caldera (Fig. 2) makes precise age control difficult to establish for the veins studied. Field relationships indicate that the veins inside the caldera and along the structures of the caldera margin are younger than caldera subsidence. These relationships are that veins crosscut (and therefore postdate) the fault rocks of the ring fault and ring dikes and mineralize faults in the central resurgent dome. The small veins in the granite outside the caldera, beyond the major structural margin faults, could possibly be older than subsidence, but we consider this unlikely because they are only found within ∼2 km of the caldera margin. ⁴⁰Ar/³⁹Ar dating (Bove et al., 2001) identified two episodes of mineralization after subsidence of Lake City caldera. The first episode is the same age as (or within error of) caldera collapse and resurgence and is represented by dated mineralization at Ute-Ulay mine (22.9 Ma) and at Red Mountain (23.1 Ma) (Fig. 2, Bove et al., 2001). The second is several million years younger, represented by dated mineralization at the Black Wonder mine (16.5 Ma), ∼800 m outside the calderas southern margin (Fig. 2). This hydrothermal episode may have been driven by heat from small rhyolite intrusions of similar age (Bove et al., 2001). We have not been able to definitively correlate the veins in this study with either of these two ages, due to the lack of a distinctive mineralogy. However, it is clear that hydrothermal fluids have utilized caldera structures, and the spatial association with resurgent intrusion suggests strongly that the majority of the mineralization is associated with resurgence. We cannot preclude the possibility that some of the mineralization is younger, and previous research in the SJVF has shown that caldera structures often control the location of hydrothermal systems that are millions of years younger (Lipman et al., 1976; Lipman, 1992).

Mapping and scanlines show that vein orientations near the caldera margin (Figs. 5B and 7B) tend to be well clustered around the structural margin orientation (apart from scanline W2, which is more variable; see Table 3), while veins farther away from the caldera margin mostly follow the regional northeast trend (Fig. 8B). The majority of veins near the caldera margin are in a four- to ten-meter-wide zone of increased vein density. The veins in the ring dike at Red Gulch may have formed by extensional reactivation and then infilling of dike-parallel shear fractures related to emplacement along the caldera margin (Fig. 7B). The veins near the southwestern caldera margin are most abundant in the highest strain fault rocks. The veins in these fault rocks follow a trend parallel to that of the caldera margin (E-W), rather than the regional NE-SW trend, showing that the structures related to caldera collapse are the primary control on fluid flow in this part of the caldera. The intense alteration, increased vein density, and hydrothermal oxygen-isotope signature of the caldera margin (Larson and Taylor, 1986b) are consistent with a high degree of fluid-rock interaction along the discontinuity. Where the caldera margin cuts across the regional northeast-trending structural grain, as at the western side of Lake City caldera (Fig. 7A), it creates additional fault and fracture intersections and potential for higher permeability (cf. Caine et al., 1996; Heap and Kennedy, 2016).

Scanlines across smaller faults inside the caldera do not always show such a clustering of orientations (e.g., circular variances of 0.48 and 0.71 for scanlines I4 and I5; see Table 3) or vein density close to the fault (Fig. 8B). Unlike scanlines SW1 and SW2, the fault cores of the faults sampled by intracaldera scanlines are thin (all one meter or less across) and almost entirely replaced or cemented by vein material (mostly quartz). The majority of veins sampled on the intracaldera scanlines are therefore in the damage zone of the faults, rather than the fault core as in scanlines SW1 and SW2. This is possibly why vein orientations are more variable on intracaldera scanlines. In the damage zone around the faults, fluids may have utilized preexisting, variably oriented, fractures formed during cooling and structural doming of the tuff and syenite.

The limited vein kinematic data available suggest that the southwestern caldera margin may have been subject to N-S extension when the hydrothermal system was active (Fig. 6). This contrasts with the NW-SE extension on the resurgent dome, forming NE-SW–trending normal faults (Figs. 2 and 8) and the E-W–extension direction suggested by numerous dilational veins on the western margin (Fig. 7). It has been suggested that Lake City caldera formed during the onset of regional E-W extension in the western United States, which formed the Rio Grande rift and Basin and Range system (Bove et al., 2001); however, there is no geologic evidence for regional extension prior to 21 Ma (Ingersoll, 2001). The range in orientations of extensional veins at Lake City caldera suggests that the stress field at the time of hydrothermal activity was variable. The regional stress cannot be determined without more widespread and robust vein kinematic data and age control, i.e., whether veins in the southwestern, western, and central caldera were contemporaneous. It can be concluded that spatial and/or temporal perturbations in the stress field were important in facilitating vein formation. The variable stress field at Lake City caldera may have been due to a complex interplay between regional stress, resurgent uplift, and extension in the apical graben in the center and on top of the resurgent dome.

Lithology has an important influence on fluid flow in hydrothermal systems (Wood et al., 2001; Rosenberg et al., 2009; Bignall et al., 2010; Farquharson et al., 2015). The physical properties of rocks can determine whether a lithology promotes or hinders fluid flow and whether fluid flow is along discontinuities or distributed through the rock matrix (Bignall et al., 2010; Heap and Kennedy, 2015; Siratovich et al., 2016). Rocks that have high strength deform brittlely to form discrete fractures that increase permeability (Rowland and Simmons, 2012). Repeated fracturing can sustain permeability after fractures seal due to mineralization. Low-strength rocks are only able to sustain fluid flow if they have a high enough intrinsic permeability, unless the processes of alteration and/or silicification increase the rock strength to a high enough level that it can fracture (Henneberger and Browne, 1988; Rowland and Simmons, 2012). In most modern geothermal systems, fluid permeability is predominantly facilitated by fractures rather than matrix or intergranular flow (Cas et al., 2011; Rowland and Simmons, 2012; Vignaroli et al., 2015; Siratovich et al., 2016). Rock strength testing has not been conducted on the lithologies at Lake City caldera; however, qualitatively one can observe that the intracaldera breccias are more porous and weaker than the granite, welded tuff, intrusive rhyolite, and syenite. This is one of the reasons why at the map scale of tens to hundreds of meters there are fewer veins in the intracaldera breccia near the caldera margin than the adjacent granite (Fig. 5A). In strong rock types near the caldera margin, such as the granite, the deformation associated with ring fault displacement will be important for creating fractures and facilitating fluid flow through the otherwise impermeable rock.

Lake City caldera and its structural margin are not uniformly altered or mineralized (Fig. 2). Understanding why some areas were subject to hydrothermal fluid flow and not others is what is important for better exploiting caldera-hosted geothermal and mineral resources. From this study at Lake City caldera, we have identified three main factors that influence fluid flow: (1) the spatial relationship of lithologies inside and adjacent to the caldera; (2) the density of discontinuities (faults and fractures); and (3) the orientation of discontinuities. Favorable combinations of these factors promote fluid flow. Table 1 shows that hydrothermal fluids can be hosted in a variety of caldera-related settings. We suggest that the favorable combination of the three factors promote hydrothermal fluid flow, and these can occur in any part of a caldera.

The arrangement of lithologies inside and near the caldera has important implications for hydrothermal fluid flow. A key requirement for any geothermal system is a heat source. In the Lake City system, the caldera-related magmatism was the heat source (Larson and Taylor, 1986b). The areas near the syenite and monzonite intrusions were, therefore, near a heat source, and are generally altered and contain mineralized faults (Figs. 2 and 8A). Similarly, parts of the ring fault containing rhyolite are always associated with alteration, whereas many parts of the ring fault without intrusions are unaffected by alteration (Fig. 2). Little alteration, or veining, is visible in the southeast part of the caldera interior where no intrusive rocks are exposed, and there is no evidence of structural pathways.

The arrangement of lithologies is also important for fluid-flow localization. If the caldera margin discontinuity consists of undeformed, welded intracaldera tuff in contact with undeformed country rock, then the margin may not exhibit higher permeability than most of the caldera fill. If the margin consists of poorly to moderately consolidated landslide breccias in contact with undeformed country rock, then fluid flow may be localized through the more porous breccias but be facilitated by matrix flow rather than through fractures or veins. Oxygen-isotope studies of the altered rocks at Lake City have found that the stratigraphically lower elevation intracaldera landslide breccias have higher water-rock ratios than adjacent welded ignimbrite (Larson and Taylor, 1986b). This suggests that breccia layers were permeable fluid pathways. The brecciated granite facies near the caldera ring fault would also have likely been more permeable than the unbrecciated granite due to the increased porosity and the granulated matrix. The permeability structure of the caldera margin will in part depend on the extent to which these fault rocks are preserved near the ring fault. Lithology has been interpreted to be an important factor in other caldera-related geothermal systems worldwide despite there being few published examples. In the Taupo volcanic zone of New Zealand, primary lithology has been recognized as an important control on hydrothermal fluid flow in the Ohaaki and Mokai geothermal systems (Bignall et al., 2010; Rissmann et al., 2011). At Ohaaki, the location of lava domes influences the distribution of permeable zones due to both the primary lithologies formed during dome emplacement and the response of these lithologies to faulting, fracturing, and sealing by hydrothermal minerals (Rissmann et al., 2011).

Discontinuities (i.e., fractures and faults) may provide permeability through rock masses with low intrinsic intergranular permeability (Sagar and Runchal, 1982; Evans et al., 1997). An area of the caldera setting that has a higher density of these discontinuities will, therefore, be more conducive to hydrothermal fluid flow. At Lake City caldera, this occurs at two different scales. At the wider map scale, the center of the caldera is dissected by numerous small-displacement normal faults (Fig. 8A), with >3 faults/km², over most of the resurgent dome, up to ∼10 faults/km². By comparison, the southeastern part of the caldera, between Sunshine Peak and Grassy Mountain, has <1 fault/km² and is the least altered part of the caldera (Fig. 2). The high density of faults (>3/km²) in the center of the caldera meant that fluid could readily flow through the welded Sunshine Peak tuff. Individually each of these faults lacks a wide “halo” of fractures around it (usually less than two meters wide), and many of these fractures (which are now infilled to form veins) are of variable orientations and unrelated to the damage zone of the fault itself (Figs. 8B and 9). This contrasts with the wide (>50 m) zone of veins in the fault core of the caldera margin fault (Fig. 5B). Because there are many faults (>3/km², compared to <1/km² in the least altered part of the caldera) inside the caldera, there is still good permeability through the rock mass, even though each individual fault is providing a narrow (<2 m) corridor for fluid flow. Resurgent domes with a high density of normal faults have been interpreted to be important for hydrothermal fluid circulation in other calderas, such as Creede (Goff and Gardener, 1994; Bethke, 2001) and Valles in the United States and Snowdon in the UK (Reedman et al., 1985) (Table 1). In contrast to the interior of Lake City caldera, the caldera margin faults and ring dikes have wider zones of increased fracture and/or vein density up to >60 m thick (Figs. 5B, 7B, and 9B). These large and highly veined zones are restricted to only a few areas along the caldera margin (e.g., Shelf Road and Red Gulch, Figs. 5 and 7) where a large halo of subsidiary fractures exhibits enhanced fluid flow.

The orientation of faults and fractures is a third important factor in controlling hydrothermal fluid flow. Certain fault and fracture orientations are more likely to intersect and promote permeability in relation to the stress field (Jolie et al., 2015). Discontinuities that have a dilational tendency (greatest when perpendicular to σ₃) and/or are critically stressed (ratio of shear stress to normal stress is greater than the critical friction coefficient of the rock) are generally most permeable (Barton et al., 1995; Jolie et al., 2015). The wide variety of orientations of extensional veins throughout Lake City caldera suggests that the orientation of principal stresses may be variable and difficult to predict in resurgent calderas. Additionally, fault and fracture networks that contain higher numbers of intersecting discontinuities, i.e., that have a higher connectivity, will be more permeable than those that do not (de Dreuzy et al., 2001; Faulds et al., 2011). The creation of discontinuity intersections is dependent on both the density and orientation of discontinuities (Faulds et al., 2011; Jafari and Babadagli, 2012; Rowland and Simmons, 2012). In Lake City, regional structure (NE-SW) interacts with subsidence and resurgent structures during postcollapse magmatism. On the western side of the caldera (Fig. 5B), the margin is almost perpendicular to the prevailing NE trend of the Eureka Graben (see also Fig. 2), and rhyolite intrusion occurs along the ring fault. At the outcrop scale, the halo of smaller veins near the margin of the ring dike closely matches the N-S orientation of the caldera margin itself (Fig. 7B), so that fracture intersections would be enhanced with the variably oriented veins outside the caldera. The southwestern portion of the caldera contains fault and/or vein intersections between the ring fault, splays off the ring fault, and mineralized preexisting dikes (Fig. 5A), but no intrusions are present. These intersections provide an additional range of structural orientations (NW-SE intersecting with E-W) associated with the ring fault, and the veins are evidence for the large volumes of fluid they transported. The sudden change in orientation at this corner in the southwestern part of the caldera is more likely to produce fault and fracture intersections than if the orientation changed gradually as a more classic subcircular ring fault. Inside the caldera, resurgent structures are typically subparallel (as are reactivated regional structures); however, slight differences in orientation and the high density (>3 faults/km²) of faults still produce intersections (Fig. 8). Favorable discontinuity orientations are therefore any that form an intersecting “mesh.” Along the caldera margin, this is most likely at ring fault corners or where the ring fault intersects regional structural grain (e.g., the N-S–oriented Lake City margin intersecting the NE-SW Eureka Graben). Fault intersections between preexisting regional faults that accommodated caldera collapse and resurgence have been interpreted to be important in localizing hydrothermal fluid flow in Long Valley caldera (Suemnicht and Varga, 1988), similar in principal to interactions interpreted in this study along the southwestern and western margin of Lake City caldera. Rytuba (1994) also emphasized the importance of intersections between caldera and regional structures in localizing ore deposits. The difference in the orientation of veins in large caldera margin faults versus the damage zone of smaller intracaldera faults may also affect permeability anisotropy. The veins surrounding a small intracaldera fault have variable orientations (Fig. 8B), which may mean the fault is equally permeable in all directions. However, on the caldera structural margin where most veins are parallel to the ring fault (Figs. 5B and 7B), permeability may be higher in the along-fault direction than across the fault. This is because fluid pathways will be more tortuous in the across fault direction, and more tortuous fracture networks result in lower permeability (Heap and Kennedy, 2015).

Faults are not usually effective conduits for fluid flow along their entire length, and fluid flow is compartmentalized to specific portions of a fault. In non-caldera settings, causes of this compartmentalization include differing sizes of fault cores and damage zones, intersections with other faults, and other structural complexities such as accommodation zones (Caine et al., 1996; Rowland and Sibson, 2004; Rowland and Simmons, 2012; Vignaroli et al., 2013; Maffucci et al., 2015). The compartmentalization occurs on a variety of scales. In the Taupo volcanic zone (TVZ) of New Zealand, the Paeroa fault system has discrete geothermal systems (Waikite, Te Kopia, and Orakeikorako) along its length separated by several kilometers (Kissling et al., 2015). The geothermal systems are sited where structural factors are particularly favorable for permeability; at Orakeikorako, there is a high density of small normal faults linking the larger Paeroa and Whakaheke faults; while Te Kopia is sited where the deformation zone associated with the Paeroa fault is particularly wide, due to extension of the fault footwall (Rowland and Sibson, 2004; Rowland and Simmons, 2012). Similar compartmentalization of geothermal fluid flow occurs along the normal fault systems of the Dixie Valley in Nevada. Extensive geothermal drilling has revealed complex patterns of faults and fractures that result in similarly complex permeability pathways in what had previously been assumed to be a large, simple normal fault (Blackwell et al., 2009). The Lake City caldera margin is not altered and mineralized along its entire length; fluid flow was focused in discrete areas tens to thousands of meters apart (Figs. 2, 5A, and 7A). This compartmentalization may be due to similar factors as those discussed above for tectonic faults. However, caldera margins may be even more likely to have large differences in permeability along their length due to added complexities that are not as common in non-caldera–related settings. These added complexities include large-scale landsliding and destruction of fault rocks (due to the large displacement accommodated in a short amount of time) and common modification of the structural margin by syndeformational and postdeformational magmatism. Therefore, the heterogeneous nature of calderas and their margins requires a conceptual framework that identifies the character and spatial relationship of factors that enhance fluid flow.

The parts of Lake City caldera that are most altered can be shown to have combinations of the three important factors: (1) favorable lithologies, (2) high densities of faults and fractures (e.g., >3 faults/km² in the center versus <1 fault/km² in the southeastern part), and (3) favorable orientations and/or intersections of faults and fractures (e.g., ring fault crosscutting regional structural grain). Hydrothermal fluid flow in the center of the caldera is aided by the close proximity to the syenite pluton heat source and a high density of small-displacement normal faults. The southwestern corner of the caldera has a high density of faults due to the structural complexities of accommodating caldera displacement at the caldera corner, and these complexities create fault intersections that enhance permeability. The western margin of the caldera is close to a heat source in the ring dike and cuts across the northeast-trending Eureka Graben, creating fault intersections that are highly permeable. Figure 10 is a conceptual cartoon of a caldera that shows how these factors may interact to explain why some areas of a caldera promote fluid flow while others do not. It must be emphasized that these three factors are not the only important variables. While these factors are important at Lake City, other factors may be as important in other systems worldwide. A similar general model for hydrothermal fluid flow in caldera settings has been proposed by Rytuba (1994), who emphasized extracaldera deformation, intersections with regional structures, and high structural complexity as favorable for the localization of ore deposits. Our conceptual model for Lake City caldera is similar to that proposed by Larson and Taylor (1986b); their model comprised a convecting hydrothermal system that utilized the ring fault and permeable breccias to transport fluid from depth to the shallow resurgent dome. However, we recognize a wider zone of caldera margin deformation and emphasize the importance of discontinuity intersections and destruction of fault rocks during subsequent gravitational collapse events to explain the discontinuous nature of veins and alteration along the caldera margin. This study has also added a quantitative understanding of the width of the fluid transport corridor around smaller-displacement (<150 m) faults, which is usually less than five meters (Fig. 8B), versus the caldera margin, which is more than 50 m across (Fig. 5B).

Detailed field work at Lake City caldera revealed evidence for fossil fluid pathways at Lake City. Evidence for fluid pathways is spatially highly variable and dependent on three key geological factors: (1) favorable lithologies, such as permeable horizons and proximity to intrusion heat sources, (2) a high density of faults and fractures, and (3) favorable fault and fracture orientations that enhance the likelihood of permeable fault and fracture intersections.

The caldera interior has favorable conditions for fluid flow due to proximity to large shallow resurgent intrusions (Fig. 8A). Permeable lithologies such as breccias are present but less common than near the caldera margins. Along the caldera margin, rhyolite ring dikes locally provide a source of heat, and widespread landslide and fault breccias are favorable sites for porous fluid flow (Figs. 5 and 10). Coherent fault rocks on the structural margin also supported abundant fractures (Fig. 5B). The ring fault itself has been destroyed in certain areas due to subsidence and related landsliding (e.g., Fig. 7A). Parts of the caldera margin that have been subject to more landsliding may have less fracture-controlled permeability through landslide breccias than areas where the structural margin and associated fault rocks are still intact (Fig. 10).

A density of >3 faults/km² is important in the center of the caldera and at the southwestern margin. The center of the caldera has a high density of small-displacement (<150 m) faults (Fig. 8A), which individually have narrower zones of veining than along the caldera margin (Fig. 8B). The high density of faults in the caldera center allows for fault intersections to occur, even though most faults have a northeast orientation (Fig. 8A). The southwestern caldera margin consists of joined rectilinear segments. Newly identified structures in granite basement outside the caldera reveal a ∼300 m zone of deformation, wider than previously thought. The outer faults utilize preexisting dike-related weaknesses to accommodate caldera subsidence (Fig. 5A). Transects across the ring fault show it has a high vein density in the fault core (Fig. 5B). Fractures in the granite accommodated fluid flow outside of the caldera margin discontinuity.

The orientation of veins close to the caldera margin discontinuity generally matches that of the main ring fault (Figs. 9A and 10). This is the case even in areas where there is a strong regional structural grain almost perpendicular to the ring fault (Figs. 7 and 10). At the western caldera margin, the intersection of the regional northeast-trending structural grain with the north-trending faults and fractures of the caldera structural margin favors the creation of permeable intersections (Fig. 7A). The structural complexities of the corner created at joined rectilinear segments at the southwestern margin also favor permeable intersections (Fig. 5A). The combination of a high density of faults and slight differences in orientation allow the creation of fault intersections in the caldera center, even though there is a strong preferred northeast orientation (Fig. 8A).

Consistent with accounts at other calderas (Table 1), our results show that both caldera margins and resurgent intrusions in caldera centers provide areas for favorable fluid flow. We establish that at Lake City the proximity to the intrusion and high density of small faults favors fluid flow within the caldera, whereas the brecciated lithologies, structural density, and structural intersections favor fluid flow at the margins.

We acknowledge Jim Cole for useful feedback on the manuscript; Cameron Windham, Jordan Lubbers, Matthew Hoffman, and Annie Waterbury for help with field work; Patrick Kelley and Cionnaith O’Dubhaigh for logistical help in Lake City; Josephine Hicks for help with scanline data entry and interpretation; and Jonathan Davidson for help with kinematic interpretation. Thomas Garden benefited from a GNS Science Ph.D. Scholarship from the Geothermal Resources of New Zealand Project Water-Rock Interactions and funding from the Mercury Energy–sponsored Source to Surface Program at the University of Canterbury and the University of Canterbury Department of Geological Sciences Mason Trust fund. Reviews from Guido Giordano, Eoghan Holohan, an anonymous reviewer, and the Editor Shanaka de Silva greatly improved the manuscript.