Complex crater fields formed by steam-driven eruptions: Lake Okaro, New Zealand

Volcanic eruptions triggered by explosive vaporization of water are common and locally hazardous phenomena. They are typically termed steam-driven eruptions (Mastin, 1995; Thiéry and Mercury, 2009; Montanaro et al., 2016c). Such eruptions eject large (meter-sized) clasts on ballistic trajectories, generate highly energetic steam-rich density currents (surges), and expel wet jets of poorly sorted rock debris (Jolly et al., 2010; Breard et al., 2014; Maeno et al., 2016; Fitzgerald et al., 2017; Strehlow et al., 2017). The main hazard from steam-driven eruptions is their unheralded sudden onset, typically having little seismic or other warning (Barberi et al., 1992; Hurst et al., 2014). More than 200 such eruptions have occurred over the last three centuries, causing thousands of deaths (Mastin and Witter, 2000). The most common steam-driven eruptions can be classified as either phreatic or hydrothermal eruptions. Both of these eruption types do not directly involve or disrupt magma, but the power of water/gas expansion explosively fragments wall rock and ejects fragments upward and outward (Browne and Lawless, 2001).

Here, we follow Stearns and McDonalds’ (1949) definition of phreatic explosions as being caused by flashing of groundwater and surface waters to steam due to the sudden arrival of heat and gas from intruding magma (or magmatic fluids). The flashed steam overpressurizes the bottom of an aquifer, but the aquifer is likely to breach near a horizontal discontinuity, such as a textural break or capping/sealing layer; thus, a simultaneous top-down rarefaction wave travels within the aquifer, and a “bottom-up” pressure wave in the overlying rock results in fragmentation and ejection of lithic clasts. Phreatic eruptions do not require the presence of a geothermal system and thus may eject fresh, unaltered rock. Deposits from a phreatic eruption should show a vertical succession of ejecta from deep sources overlain by ejecta from shallower sources.

According to Mastin (1995) and Browne and Lawless (2001), hydrothermal eruptions are triggered within a geothermal area by a variety of processes that cause the sudden decompression and flashing of water that is already metastable and near boiling. Hydrothermal eruptions typically progress from a near-surface rupture downward (McKibbin et al., 2009). This mechanism is energetically very favorable, and a hydrothermal eruption can start within a meter or so of the ground surface below a very thin cap. After initial excavation, pressure within the deeper geothermal reservoir is reduced, and a flashing front and rarefaction front move progressively downward, followed by the boiling front. Water present in joints or cracks adjacent to the developing crater may also flash to steam as pressures reduce suddenly, widening the vent. This process may occur in both brittle or unconsolidated materials (Browne and Lawless, 2001; Galland et al., 2014; Montanaro et al., 2016b). Deposits typically contain abundant hydrothermally altered components and may record a progressive deepening of explosion locus through an inversion of the pre-explosion stratigraphy.

Steam-driven eruptions last from seconds to hours (Browne and Lawless, 2001; Jolly et al., 2014) and produce craters from a few meters up to hundreds of meters in diameter, with depths from few to several hundred meters (Muffler et al., 1971; Browne and Lawless, 2001; Morgan et al., 2009). Deposits produced by these explosions are generally of low volume (<10⁵ m³) and restricted to within hundreds of meters to a few kilometers from crater margins. They are typically very poorly sorted, matrix-supported breccias (Muffler et al., 1971; Mastin, 1995; Browne and Lawless, 2001; Breard et al., 2014). Steam-driven eruptions of all types occur in diverse volcanic and sedimentary rocks, showing a range in grain size, competence, alteration, fracturing, and bedding. All these factors influence deposit distribution and crater form. The geological and stratigraphic setting of an eruption site is also an important factor in interpreting eruption dynamics (Browne and Lawless, 2001; Morgan et al., 2009; Breard et al., 2014; Valentine et al., 2015a; Montanaro et al., 2016b).

Steam-driven eruption craters are common in many volcanic terrains or areas of high heat flow, such as in New Zealand (e.g., Champagne Pool crater at Wai-o-tapu; Hedenquist and Henley, 1985), Indonesia (e.g., Sinila and Sigludung craters in the Dieng Plateau; Allard et al., 1989), Japan (e.g., craters on the Ontake volcano summit area; Maeno et al., 2016), Greece (e.g., Stefanos crater on Nysiros; Marini et al., 1993), El Salvador (e.g., craters in the Agua Shuca thermal area; Handal and Barrios, 2004), and the United States (e.g., Mary Bay crater complex in Yellowstone National Park; Morgan et al., 2009). In all these cases, craters were excavated within a wide range of host-rock lithologies having different alteration states, strengths, and permeabilities (Maeno et al., 2016; Montanaro et al., 2016a; Heap et al., 2017). These factors produced a typically complex crater morphology. Craters from steam-driven eruptions may preserve evidence of multiple explosions that migrated laterally (Nairn and Wiradiradja, 1980; Cole et al., 2006; Morgan et al., 2009; Breard et al., 2014). Evidence includes: lines or clusters of closely spaced craters; a single large crater with scalloped margins and/or a complex shape; and nested craters with different depths (e.g., Scott and Cody, 1982; Morgan et al., 2009; de Ronde et al., 2015; Montanaro et al., 2016b). Detailed mapping of the ejecta blanket, as well as ballistic ejecta analyses around the crater area, may reveal additional evidence of multiple explosion epicenters during a single active episode (Nairn, 1979; Kilgour et al., 2010, 2019; Breard et al., 2014; Lube et al., 2014; Maeno et al., 2016; Montanaro et al., 2016b).

In this study, we assessed the effects of variations in the eruption-generation mechanism (bottom-up vs. top-down), the host-rock lithology, and the impact of successive eruptions on explosion dynamics and crater formation processes, using the example of Lake Okaro in New Zealand, located within the Taupo volcanic zone (Fig. 1). Lake Okaro has a subrectangular shape, and the surrounding breccia (Cross, 1963; Hedenquist and Henley, 1985) indicates that it may have resulted from multiple explosions. Hardy (2005) recognized at least three breccia units produced by an initial phreatic eruption at the southern end of the lake; this eruption was followed by a series of hydrothermal explosive events, enlarging the crater to its final shape. Some of the later hydrothermal eruptions were larger than the initial phreatic event. The eruptive activity that formed Lake Okaro is representative of many other complex cases of repeated steam-driven eruptions in New Zealand (e.g., at Rotokawa, Waimangu, and Wai-o-tapu geothermal fields), as well as many other similar areas around the world (e.g., large hydrothermal eruption craters in Yellowstone National Park, United States; Browne and Lawless, 2001; Morgan et al., 2009). Here, we examined the breccia deposits surrounding Lake Okaro and investigated the lake bathymetry to: (1) revisit the eruptive triggers, and (2) explore the relationship between eruption dynamics, including crater-forming processes, and the host-rock properties during a series of repeated steam-driven explosive events.

Lake Okaro is located within the central Taupo volcanic zone (Fig. 1), north of the Wai-o-tapu geothermal field. This area is dissected by the NE-striking Ngapouri and Rotomahana fault systems, which feed fluids to numerous geothermal fields (Hedenquist and Henley, 1985; Nairn et al., 2005). Nairn et al. (2005) suggested that these fault/fluid pathways were reactivated during magma rise into the neighboring Okataina caldera, triggering the A.D. 1315 effusive and explosive, rhyolitic Kaharoa eruptions at Tarawera, as well as many of the steam-driven explosive eruptions within the Wai-o-tapu geothermal field. Contrasting with many hydrothermal craters in this area, Lake Okaro is not directly above a major fault system; it lies 0.5 km northwest of the nearest surface faults (Ngapouri and Rotomahana faults; Figs. 1 and 2). An inferred fault is mapped on the western side of the lake, and rivers running parallel to the local fault orientations on the eastern lake shore could indicate the presence of older faults and fractures across the whole Lake Okaro area (inferred faults in Figs. 1 and 2).

The stratigraphy below Lake Okaro (Fig. 2) was constructed from nearby field observations and borehole logs. The deepest units recognized in drill cores (Nairn, 2003; Hardy, 2005) include an undescribed siltstone unit overlain by the Rangitaiki Ignimbrite (RI in Fig. 2). The Rangitaiki Ignimbrite is a poorly to moderately welded, dark-gray, crystal-rich tuff (Nairn, 2002). Facies include coarse ash- to pumice-flow tuffs, pumice breccias, and ash-to-lapilli-fall deposits. This ignimbrite is >100 m thick on the caldera rim at Lake Rotomahana, 3 km from Lake Okaro (Fig. 1), but it thins to only 7–39 m thick in boreholes near Lake Okaro (Nairn, 2002). The thinning likely reflects the pre-eruptive topography from displacement associated with blind fault branches of the Rotomahana fault system (inferred faults in Figs. 1 and 2). A 9-m-thick tuff-breccia, rich in accretionary lapilli, lies above the Rangitaiki Ignimbrite (Nairn, 2003). This unit is correlated with a similar breccia near Haumi Stream (Fig. 2), interbedded with a bedded lapilli-fall deposit (Rotoehu Ash; RA in Fig. 2; Nairn and Kohn, 1973; Leonard et al., 2010). The Earthquake Flat Formation overlies a sharp nonerosional break above the Rotoehu Ash (EFF in Fig. 2; Nairn and Kohn, 1973). The Earthquake Flat Formation is a weakly compacted, crystal-rich, lapilli-bearing ash, with coarsely vesicular pumice blocks; coarse (>5 mm) quartz, plagioclase, hornblende, and biotite crystals and crystal fragments are also present (Molloy et al., 2008; Leonard et al., 2010). In boreholes close to Lake Okaro, the Earthquake Flat Formation is ∼50 m thick (Figs. 1 and 2; Nairn, 2003; Hardy, 2005), and it is overlain by 0.5–2-m-thick ash-to-lapilli-fall deposits of the Waiohau Formation erupted from the Okataina volcanic complex (Nairn, 2002; Speed et al., 2002; Leonard et al., 2010; WF in Fig. 2). Above the Waiohau Formation, there is a series of younger tephras, including the 1.8 ka Taupo Pumice Formation (TPF in Fig. 3). A 1–3-cm-thick ash-fall deposit from the ca. A.D. 1315 Kaharoa eruption lies within a paleosol immediately below the Okaro breccias (Lloyd, 1959; Cross, 1963; Hedenquist and Henley, 1985; Nairn et al., 2005).

We described the Okaro breccia deposits in several exposures around the lake (Figs. 1 and 3–7; Figs. DR2–DR5¹) and sampled the ash and lapilli-rich matrix, as well as blocks representative of the main lithologies, to be used for component and grain-size analyses. The percentage of blocks with respect to the breccia matrix was qualitatively assessed for the investigated sections. Examination of macroscopic alteration features, for example, color differences from whitish (freshest) to orange, yellow, reddish, and greenish (most altered), as well as the occurrence of alteration halos and silicified crusts and veins, indicated the relative degree of alteration within the investigated breccias. Field observations, in concert with the lake bathymetry (described later), were then used to reconstruct the eruptions scenario.

Three main breccia units, Okaro Breccia I, II, and III, were distinguished based on grain-size, color, and lithology/componentry (Figs. 3–6), and they are grouped within the Okaro Breccia Formation. Cross (1963) and Hedenquist and Henley (1985) mapped the Okaro Breccia Formation out to ∼1 km from the lake shoreline, with one lobe extending ∼1.5 km to the east (Fig. 1). The whole formation is ∼13 m thick near the crater rim, but it rapidly thins to 2–3 m at ∼250 m from the eruptive center.

On the western lake shores (location 1 in Fig. 1), the Okaro Breccia Formation overlies a 1-cm-thick fall unit from the A.D. 1315 Kaharoa eruption and a series of tephras from the Taupo and Okataina calderas, capping the Earthquake Flat Formation ignimbrite (Figs. 3, 4, and 7; Fig. DR1). A thin, pale brown paleosol caps the Okaro Breccia Formation, which is in turn covered by the Rotomahana Mud, a distinctive deposit from the phreatomagmatic phase of the 1886 Tarawera eruption (Fig. 7; Figs. DR3, DR4, and DR6; Nairn, 1979).

The basal Okaro Breccia I is orange to yellow brown and dominated by an ash-rich matrix, supporting subrounded to angular lapilli and blocks (maximum diameter 100–150 mm; Fig. 4; Fig. DR1). The block fraction (>64 mm) consists of two main types of slightly altered to unaltered lithologies: (1) white to orange-stained pumice-rich tuff and fresh crystal-rich tuff, both with abundant lithic clasts, biotite, and quartz grains, and (2) fine-ash tuff containing abundant biotite, pumice, accretionary lapilli (0.5–1 mm sized), and lithic clasts in a pale-gray to yellow matrix. The first block lithology is predominant in the breccia (>60 modal %) and is derived from the Rangitaiki Ignimbrite, whereas the second, less abundant type (>35 modal %) is derived from the Rotoehu Ash. Rare (<5 modal %) white pumice blocks, rich in biotite and quartz, derived from the Earthquake Flat Formation, are also present.

The matrix of Okaro Breccia I is reversely graded from ash-rich upward to greater lapilli content (reaching up to ∼30 wt%; Figs. 8A–8D). There is a persistent grain-size mode between 0 and –0.5 phi (1.4–1 mm; very coarse ash), as well as a long tail (from ∼30 to ∼50 wt%) extending from 1 mm to fine ash (<0.063 mm; Figs. 8A–8D). The matrix consists mainly of slightly altered to unaltered fragments of the same ignimbrite, tuff, and pumice as the blocks, together with loose biotite and quartz crystals (Figs. 8A and 8B). Ignimbrite and tuff clasts are orange-stained and subangular to angular, collectively making up ∼50 modal % of the particles (Figs. 9A and 9B). Another 30–40 modal % of the matrix particles are white to yellow pumice, which are mostly angular, highly vesicular, and rich in quartz and biotite. They are most abundant in the upper part of the breccia. Free crystals make up >10 modal % of the matrix and are mostly quartz, particularly in its middle to upper portion. The matrix particles commonly show a yellow alteration patina. Additionally, rare but distinctive red-stained, tuff-like particles occur within the matrix at the base of the unit (Figs. 9A and 9B).

The Okaro Breccia I is thickest in the western (1 m) and southwestern (2.5 m) quadrants around the lake. It is not present to the north, and there are no outcrops of this member in the south and east (sections 1, 3, and 10; Fig. 7).

In the southwestern sector of the lake (location 3; Fig. 1), the Okaro Breccia II is separated from the underlying Okaro Breccia I by a thinly laminated layer (1–10 cm thick) of reworked and waterlain ash and a pale brown paleosol (Figs. 5A–5B). However, at other locations (e.g., 1 in Fig. 1), no erosional surfaces or sedimentary layers are present between the two units (Fig. 4). In the northern and northeastern sectors, the Okaro Breccia II directly overlies the Kaharoa ash (Fig. 7; Figs. DR2 and DR3). The Okaro Breccia II is the most extensive deposit of all deposits mapped around the lake, covering all sectors. The breccia is at least 10 m thick in the southern and northern sectors of Lake Okaro, 6.5 m thick to the southwest, and 1.5 m thick to the east (Fig. 7).

The Okaro Breccia II shows a vertical stratification with diffuse contacts between beds (Figs. 3 and 7). The basal bed is yellowish to gray, dominated by fine-grained matrix-supported subrounded to angular lapilli. The massive central bed is brown to yellow-brown and matrix- to clast-supported, and it contains angular to subrounded lapilli to coarse blocks (maximum diameter 400–600 mm; Figs. 5C–5D; Fig. DR4). The uppermost bed is brown to orange and dominated by a matrix-supported subrounded to angular lapilli (Fig. 6). The full sequence is present in the southern and western sectors around the lake, whereas to the east, northeast, and north, only the central bed is present with rare large blocks. In all the investigated sections, the central bed is characterized by distinctive altered tuff-like green clasts as blocks and/or within the matrix (Figs. 7; Figs. DR3 and DR5).

The block fraction of Okaro Breccia II consists of fine-ash (>40 modal %) and accretionary lapilli–rich tuff (<30 modal %), and biotite- and quartz-rich pumice (∼30 modal %), mainly in the massive central bed (Figs. DR4–DR6). The first block lithology is derived from the Rangitaiki Ignimbrite, whereas the second type is derived from the Rotoehu Ash. The third lithology, more abundant compared to the amount found in the Okaro Breccia I, is derived from the Earthquake Flat Formation. There is a larger proportion of blocks (up to 30–40 modal % in the massive bed) cropping out in the southern, southwestern, and western sectors (Figs. 3, 5, and 7; Fig. DR5). The tuff clasts vary from pale gray to white to greenish-white and are mostly angular, whereas the white pumice particles are rounded to subangular. In the upper bed, rare silicified vein-like clasts and brecciated tuffs are found (Fig. DR4). Rare bomb sags were identified within the basal bed (Fig. DR3).

The matrix grain-size distribution of Okaro Breccia II is coarsely skewed, having a large mode in the lapilli fraction and tails of ash (<15 wt%; Figs. 8E–8F). The central massive bed has the greatest ash content (Fig. 8F). The matrix contains mainly quartz- and biotite-rich pumice (>50 modal %), decreasing in content upward (Figs. 9C–9E). Tuff clasts are subordinate (<30 modal %), and their proportion increases from the center of the middle bed upward. Some matrix particles of all types show a yellow alteration patina. Up to 10 modal % of matrix particles are strongly altered, green, and rounded to angular in shape. These particles are limited to the upper half of the Okaro Breccia II (from the middle part of the massive bed upward; Figs. 6 and 8D–8E; Figs. DR3–DR5). Abundant quartz crystals (<10 modal %), either euhedral or broken, together with rare unidentified dark clasts (<1 modal %) occur in the matrix (Figs. 9C–9E).

The Okaro Breccia III crops out in the southern and southwestern part of the lakeside and directly overlies the Okaro Breccia II, commonly without a weathered contact or paleosol. At the type section, a thin (up to 2 cm), fine-grained, waterlain orange ash deposit separates it from the Okaro Breccia II (Fig. 6). The Okaro Breccia III is red and mostly clast-supported, containing predominantly subrounded to angular lapilli, and it is capped by modern/recent soil. The Okaro Breccia III is limited to the western and southwestern sectors around the lake, where it is 2.5 and ∼0.5 m thick, respectively (Figs. 1 and 7).

The block fraction of Okaro Breccia III mainly consists of angular fine-ash tuff (>60 modal %) and accretionary lapilli–rich tuff (>30 modal %), with a minor quantity (<15 modal %) of biotite- and quartz-rich pumice. Generally, all clast types are whitish in color and less competent than those within the other two breccias. Similar to the components of the first two breccias, the first and second block lithologies are derived from the Rangitaiki Ignimbrite and the Rotoehu Ash, respectively, whereas the pumice is derived from the Earthquake Flat Formation. Rare (<5%) altered tuff-like green blocks occur. Few large blocks (up to 30 cm) are present at the base of the breccia, whereas the upper part is lapilli-rich with rare blocks (Fig. 6). No bomb sags were identified within this member.

The matrix grain-size distribution is coarsely skewed (Fig. 8H) and is lapilli dominated (>60 wt%). The matrix consists mainly of slightly or strongly altered pumice particles (>50%), subordinate (<20%) loose crystals (euhedral and broken quartz), and strongly altered fragments of tuffs (Fig. 9F). All matrix pumice particles are rounded to angular, yellowish-white, or covered by a red alteration patina. Altered tuff-like green clasts occur only in the block fraction.

Lake Okaro has a roughly rectangular shape, is ∼650 m long and 400 m wide, and covers an area of 0.31 km² (Fig. 10). High-resolution multibeam data (1 m spatial resolution) were acquired in 2014 by the Bay of Plenty Regional Council (Figs. 10 and 11). The level of Lake Okaro has varied historically (up to 1.5 m) in response to changing rainfall patterns (Cross, 1963). The lake surface had an elevation of 413 m during the 2014 survey and showed a maximum depth of 18 m in its southern sector. We calculated a water volume of 3.9 × 10⁶ m³ using the high-resolution multibeam data.

A well-developed drainage network is present west and northwest of the lake (Fig. 1), supplying sediment and water to form fans on the lake floor. On the northwestern hillsides bordering the lake, there are numerous rills, which formed immediately after the emplacement of the Rotomahana Mud during the A.D. 1886 Tarawera-Waimangu eruption (Hardy, 2005).

The lake bathymetry allows reconstruction of the order of formation of the Okaro eruption craters. We determined three subareas of craters (Figs. 10 and 11; Fig. DR6), representing separate eruptive phases that produced the three separate breccia units. We defined these subareas based on the fact that younger eruption craters (1) cut the earlier-formed crater borders, (2) excavated into older breccias, and (3) typically showed fresher morphology (deeper, steeper walls and sharper margins). Older craters are smoothed and/or shallowed by infill of younger ejecta and/or lake sediment. Crater shapes were defined as circles or ovals fitting the distinct (or inferred) crater rims (Fig. 10; Fig. DR6). Average crater diameters and slope angles were measured, together with the depth (from the crater rim to the deepest point in the crater), and these value are listed in Tables 1 and 2.

Phase I craters. Few craters were definitively attributed to this phase because they are generally the most eroded, buried by sedimentation, and crosscut by later craters. They include the less obvious crater features located in the southern part of the lake (Fig. 11). Many of the craters, especially on the southwestern and southern sides, are represented only by remnants of their original crater walls (Fig. 10, profiles A-A′, B-B′, and E-E′). In the southeastern part of the lake, the crater borders are more evident, and they have relatively smooth floors. A NW-SE–elongated ridge in this area probably represents another remnant crater that formed early in the sequence (Fig. 10, profile E-E′). The few craters with shapes that can be extrapolated (1–3 in Fig. DR6) have diameters between 62 and 120 m, and rim-to-crater-floor depths of 1.8 and 5 m. Craters from phase I exhibit very steep walls (from 22° to 49°; Table 2) and have a low crater depth/diameter ratio (0.03–0.04). Inferred craters from phase I form a linear chain likely oriented along a fracture system (inferred Rotomahana fault in Figs. 1 and 2). There may have been craters northward of this inferred chain, but if they existed, they were later destroyed by the phase II eruption. Based on the inferred subareas, the phase I eruption involved ∼0.1 km², or one third of the current lake area.

Phase II craters. At least 20 phase II craters crosscut phase I crater areas. These craters are scattered across the lake floor (Fig. 10; Table 1). In the northern lake, large craters (4, 6, 7 in Fig. DR6) occur, having diameters between 110 and 190 m and depths between 9 and 18 m (Fig. 10, profile A-A′ to D-D′). These craters show intermediate wall slopes (from 18° to 25°) and have crater depth/diameter ratios from 0.05 to 0.12. Small sediment-covered depressions and circular craters (e.g., 5 and 8–13 in Fig. DR6) are present among, or cut, the large craters (Fig. 10, profile D-D′). These features have diameters of 47–70 m, depths between 1.4 and 2.8 m, and depth/diameter ratios of 0.02–0.06. Clusters of small, coalesced craters (14–20 in Fig. DR6) occur in the eastern and southeastern parts of the lake (Fig. 10, profile E-E′), having diameters between 33 and 80 m and depths of 1.6–3.9 m. Most of these clustered craters show low wall slopes (from 11° to 15°) and have crater depth/diameter ratios between 0.04 and 0.07. In the southwest, intermediate-sized craters (21–22 in Fig. DR6) have diameters of 77 and 92 m, intermediate wall slopes (16°–29°), and crater depth/diameter ratios of 0.03 and 0.06. All of these intermediate-sized craters cut the remnant phase I craters, as well as further excavate their crater floors (Fig. 10, profile E-E′). There may be more craters buried beneath the landslide deposits that blanket the northern part of the lake. Phase II craters cover ∼0.25 km².

Phase III craters. These craters crosscut crater walls from phases I and II, representing the most geomorphologically distinct and youngest craters occurring on the southern side of the lake (Fig. 10, profiles A-A′ and E-E′). On their eastern and southern sides, coalescing craters (25–28 in Fig. DR6) cut into a NW-SE–elongated ridge and remnants of older crater borders, both from phase I. On the northern and western sides of the phase III subareas, craters cut the rims and floors of craters from phase II (23, 24, 29 in Fig. DR6). Inferred diameters for the coalesced craters range from 53 to 110 m, and depths are between 1 m and 3 m. In general, these craters exhibit low wall slopes (from 5° to 15°) and low crater depth/diameter ratios (0.02–0.03). Craters of phase III (at least seven distinguishable) cover an area of ∼0.035 km².

Along the western and northwestern sides of Lake Okaro, several features indicate sediment inflow to the lake below creeks/valleys that drain the surrounding hills (Figs. 1, 10, and 11). A large delta occupies ∼0.05 km² on the northwest side of the lake, and one of 0.008 km² is present on the northeastern side. Stepped escarpments, 1–3 m high, on and above the largest delta indicate that this is the toe of a retrogressive slump feature (Fig. 10, profiles A-A′ and C-C′; Fig. 11). On the western side of the lake, small rockslide deposits have a hummocky topography and include scattered megablocks (up to 20 m wide and 1.5 m high) below many of the steeper crater wall embayments (Fig. 10, profile C-C′; Fig. 11).

Potential triggers of the initial explosive activity at Lake Okaro and eruptive scenarios can be inferred from several lines of evidence, including subsurface stratigraphy and tectonic structures, as well as the alteration state and stratigraphy of the Okaro Breccia Formation.

Lake Okaro lies close to the Rotomahana multithread fault system (Fig. 1; Lloyd, 1959; Hedenquist and Henley, 1985). Displacement along these faults produced a raised local topography, so that only a thin veneer of Rangitaiki Ignimbrite was emplaced, and the regional topography smoothed (Nairn, 2003). Also, a faulted volcanic sequence outcrops in the nearby Haumi Stream (Nairn et al., 2005), along with a surface fault west of Lake Okaro. Indirect evidence of further fault strands includes streams oriented parallel to the main fault trends immediately northeast of the lake (Fig. 1). Collectively, it appears that an approximately NE-SW–striking fracture zone existed at the site of Lake Okaro before the eruption (inferred fault in Figs. 1 and 2).

The phase I eruption occurred soon after the explosive phase of the A.D. 1315 Kaharoa rhyolitic eruption (volcanic explosivity index [VEI] 4; Bonadonna et al., 2005; Nairn et al., 2005). Nairn et al. (2005) suggested that basalt dikes across the region near Mount Tarawera injected CO₂ and heat to prime several steam-driven eruptions, e.g., the Champagne Pool at the Wai-o-tapu geothermal system. Fault displacements and/or channeling of magmatic gas up faults and fractures could have thus triggered the widespread explosive episodes, including Lake Okaro (Hedenquist and Henley, 1985; Browne and Lawless, 2001; Rowland and Simmons, 2012).

Okaro Breccia I includes the deepest local lithology (Rangitaiki Ignimbrite; Nairn, 2003; Hardy, 2005) as tuff clasts at the base of the unit. There are no juvenile pyroclasts, indicating that no phreatomagmatic eruption occurred. These Rangitaiki Ignimbrite clasts show that the initial explosion began at least at a depth of ∼70 m. In addition, there are very few altered clasts in Okaro Breccia I, so no significant geothermal system existed at the eruption site.

Collectively, the initial eruption of deep-seated bedrock, the absence of pervasive hydrothermal alteration of the ejecta, and the presence of faults and fractures in the area suggest that the eruption was triggered by rapid heating and overpressurization of groundwater by sudden arrival of magmatic fluids, for instance, from a dike (cf. Germanovich and Lowell, 1995; Stix and De Moor, 2018). This interpretation follows Nairn et al. (2005), who proposed that magmatic CO₂ injection from a basaltic dike generated hydrothermal eruptions at Wai-o-tapu. Nearby, a shallow dike also caused both steam-driven and phreatomagmatic eruptions at Waimangu in 1886 (Nairn, 1979). Thermodynamic models (e.g., Delaney, 1987) suggest that heat is not transferred rapidly from an ascending dike to groundwater. However, large amounts of rising CO₂ from a basaltic dike (>1000 t/d; Nairn et al., 2005) could have rapidly heated and pressurized shallow groundwater in fractures and aquifers above faults at the site of Lake Okaro. Germanovich and Lowell (1995) considered emplacement of magmatic fluids into a water-saturated permeable reservoir with two scales of permeability: low permeability (<10⁻¹⁷ m²) in the bulk rock, but high permeability around crack networks (>10⁻¹² m²). In this case, explosive-eruptive conditions are reached when the fluid of the subsidiary network starts to be heated, leading to microscale pressurization. Heat and gas cause rapid propagation of cracks (seconds to hours for fractures 10⁻³ m to 10⁻¹ m in size), leading to near boiling (Germanovich and Lowell, 1995). Extensional stresses and overpressures build up and eventually cause the country rock to fail, initiating decompression and an explosive eruption. This model is well suited for rocks with low tensile strength (≤10 MPa), such as the Rangitaiki Ignimbrite (4–8 MPa; Foote et al., 2011), which also has a bulk permeability of 10⁻¹⁶ m² (Montanaro et al., 2017).

A steam-driven event along a fracture system above a blind fault and a dike is also consistent with the linear chain of vents/craters of phase I (Fig. 11). Following phase I explosions, the elongate crater structure was filled by highly permeable ejecta. Ongoing heat transfer from the shallow intrusion may have lasted for months to years (cf. Petcovic and Dufek, 2005), heating and circulating hydrothermal fluids (cf. Rowland and Simmons, 2012). This geothermal system expanded into the shallow permeable (∼10⁻¹² m²) deposits of the Earthquake Flat Formation surrounding the phase I craters (Fig. 12B; Tschritter and White, 2014).

The new shallow hydrothermal system covered at least the area of phase II craters, extending ∼400 m north of the phase I explosion sites (Figs. 11 and 12B). The occurrence of thinly laminated lacustrine sediments between Okaro Breccia I and II (Figs. 5A and 5B) indicates a time break between these phases, with formation of a lake and related sediment accumulation. The presence of silicified sediments, silicified breccias, and abundant altered green tuff clasts, as well as large quantities of altered pumice and tuffs within Okaro Breccia II all around the lake (Figs. 9C–9E; Fig. DR4) suggests at least decades of hydrothermal activity took place before phase II explosions. In this setting, thermal and mechanical conditions at Lake Okaro were primed for producing a hydrothermal eruption, under any trigger scenario (e.g., Browne and Lawless, 2001). Possible trigger mechanisms of phase II eruption include seismic displacements and changes in surface and groundwater in the Lake Okaro area (cf. Lawless, 1988; Rowland and Simmons, 2012).

The Okaro phase II eruption was likely triggered by a seismic event that fractured the hydrothermal system and reduced the confining pressure. The decompression of hydrothermal fluids probably resulted in the formation of a boiling-erosion front that penetrated and excavated down into the hydrothermal reservoir, mostly within the <70-m-depth Earthquake Flat Formation (top-down model of McKibbin et al., 2009). The eruption continued until the rate of groundwater boiling decreased and steam expansion declined to the point where rock could no longer be ejected from the crater. The collapse of crater walls, or flooding may have further contributed to stopping the eruption (Browne and Lawless, 2001).

Another pause in the activity occurred between phases II and III, as indicated by a thin, waterlain mud deposit below Okaro Breccia III (Fig. 6), on the western side of the lake and positioned ∼3 m above the current lake level. This deposit indicates that a large lake formed in the depression produced during phase II. Hardy (2005) also suggested that the lake level was once much higher (>3 m), based on undercutting to the north and terracing to the east of the present Lake Okaro; our finding of lake sediments on the phase II crater rim confirms this hypothesis. The Okaro Breccia III mostly consists of tuff and minor pumice that appear more altered than those from Okaro Breccia II, and its extent is limited to the western to southwestern area of the lake close to the source craters (Figs. 7 and 11). Further alteration of the Okaro Breccia III clasts indicates some residual hydrothermal activity within the strongly reworked/fractured ejecta of the phase I crater area (Fig. 12C). Hardy (2005) also noted that erosion and the formation of a 50 m breach in the southeastern margin of the lake resulted in catastrophic drainage. Drainage may have caused a sudden decrease in confining pressure beneath the lake bed, which we suggest triggered rapid boiling in the surficial hydrothermal system and the onset of phase III eruption (Fig. 12C). Similar events have occurred after subtle lake drainage in other shallow hydrothermal systems, excavating top-downward (cf. McKibbin et al., 2009), and producing craters of comparable size to the phase III craters, for example, in Yellowstone and Iceland (Muffler et al., 1971; Morgan et al., 2009; Montanaro et al., 2016b).

Field studies of natural explosion craters (Yokoo et al., 2002; Montanaro et al., 2016a), and field-based explosion experiments using loose-to-compacted material at varying shallow depths (Murphey and Vortman, 1961; Goto et al., 2001; Ohba et al., 2002; Taddeucci et al., 2013; Graettinger et al., 2014; Valentine et al., 2014; Sonder et al., 2015; Macorps et al., 2016) have demonstrated that crater shape and size result from an interplay between explosion energy and scaled depth (physical depth divided by cube root of energy). In addition, preexisting craters appear to influence the ejecta jets and whether or not an explosion is able to vent (Taddeucci et al., 2013; Graettinger et al., 2014). Experimental results further indicate that the volume affected depends on the host-rock (substrate) strength (Galland et al., 2014; Macorps et al., 2016). In many natural locations, multiple neighboring craters may imply internal inhomogeneity in fluid storage within rock bodies (e.g., thermal circulation patterns that set up mineralization boundaries at their margins, or three-dimensional variations in the porosity and permeability characteristics of host rock; Lawless, 1988; Browne and Lawless, 2001; Rowland and Simmons, 2012; Montanaro et al., 2016a). At Lake Okaro, much of the hydrothermal water was contained within the highly permeable, weakly compacted Earthquake Flat Formation (Tschritter and White, 2014). In areas near Lake Okaro, this formation shows complex changes in deposit texture, incipient compaction, and gas-escape pipes (Leonard et al., 2010). These variations may extend at depth underneath the Lake Okaro area, thus setting up “pockets” of coexisting fluids with only secondary connections to the rest of the hydrothermal system. These pockets could explain the multiple craters, rather than the formation of one large crater area (Kilgour et al., 2019).

Macorps et al. (2016) set up explosion experiments in “strong substrates” (mimicking well-consolidated sediments) and “weak substrates” (mimicking unconsolidated sediments or volcaniclastic deposits). Their results suggest the following:

• (1) The host rock properties affect the crater morphology (e.g., steepness of walls), the clast type, and sizes available for ejection;

• (2) the host rock disrupted by subsurface explosions loses its original strength and has less impact on subsequent explosions, and

• (3) most of the disrupted subsurface structure is filled with ejecta.

As a first approximation, strong substrates at Lake Okaro could correspond to the deep Rangitaiki Ignimbrite and Rotoehu Ash tuffs, whereas weak substrates could correspond to surficial tephra deposits (Waiohau and Taupo Pumice Formations), or the ejecta from phases I and II. The Earthquake Flat Formation, representing the thickest and main host rock involved in the first two phases of Okaro eruptions, varies from weakly compacted to possibly slightly consolidated at depth, thus acting as a substrate with intermediate strength (Macorps et al., 2016).

The steepness of crater walls at Lake Okaro (Table 2; Fig. 10) appears to have been influenced by the presence of the Earthquake Flat Formation. Many of the craters formed during the first two phases were excavated within this formation and have steep sides (23°–49°). By contrast, craters excavated within unconsolidated tephra deposits or the ejecta deposits of phase I and II eruptions have low-angle slopes (6°–15°).

We infer that the Okaro Breccia I eruption crater distribution was dominated by the focusing mechanism of deep hot fluids in a preexisting fracture zone. Thus, the craters were elongate and aligned along this NE-SW–striking fracture system (Figs. 1 and 2). Based on the slope values and bathymetric profiles (Table 2; Fig. 10), phase II and III craters either excavated deeper into the phase I (increasing slope angles), or deposited ejecta that filled the crater (decreasing slope angles). We suggest that the phase I eruption was hosted mostly within the Earthquake Flat Formation, and it was deeply excavated along a fracture to produce a steep-sided fissure-like crater. The Okaro phase I eruption produced a breccia that was less widespread than the phase II breccia, consistent with the greater depth of initiation of the phase I phreatic eruption.

The phase II eruption produced the most widespread and thickest breccia, as well as many craters with different sizes and a broad distribution (Fig. 11). Moreover, the phase II eruption involved a wide area north of the initial eruption site, which may reflect the presence of a large hydrothermal system. The wide area may also reflect the fragmented and/or weakly compacted nature of the host rocks and the intermediate to shallow depth of the explosions within the Earthquake Flat Formation (Galland et al., 2014; Montanaro et al., 2016b).

The final phase II crater shapes (Fig. 11) suggest that a series of blasts overprinted and enlarged the craters from preceding events, while maintaining a roughly circular shape (cf. Valentine et al., 2015a). The smaller nested craters may have resulted from multiple small steam explosions (Scott and Cody, 1982; Shanks et al., 2005; Morgan et al., 2009). Phase II craters in the east and southwest mostly excavated ejecta from phase I, without extending the original crater area (Figs. 10 and 11). Their coalesced shape may reflect shallow explosion loci.

Craters produced during the phase III explosions are shallow compared to those produced by the previous eruptions, and the explosions recycled ejecta from the previous two eruptive phases. During phase III, almost no energy was spent in fragmenting the host substrate (cf. Montanaro et al., 2016b), and the excavation produced overlapping shallow structures, with shallow explosion loci.

Breccia componentry reflects the different host rock and depths of fragmentation during the three periods of eruptions. Okaro Breccia I is dominated by weakly altered and unaltered tuffs indicating an explosion locus >70 m depth, and overburden pumice from shallow levels (Fig. 9A). The deeper, consolidated units required high energy to be fragmented and ejected, whereas less energy was consumed in “unloading” or disaggregating the Earthquake Flat Formation (Alatorre-Ibargüengoitia et al., 2010; Montanaro et al., 2016b, 2016c). The Okaro Breccia II shows a complex circumcrater variation in componentry and grain size, as well as in sorting and thickness (Fig. 7). This variability reflects the contrast between undisturbed Earthquake Flat Formation excavated in the northern sector and prefragmented ejecta from phase I in the southern sector. The radial variation of Okaro Breccia II thickness may reflect instabilities in the eruptive jets, multiple different explosion positions, as well as directed jets produced by interactions with confining crater walls (cf. Taddeucci et al., 2013; Valentine et al., 2015a, 2015b). Directed jets, in particular, produce strongly asymmetrical skirts in which the thickest deposits are on the crater side opposite to the jet direction (Graettinger et al., 2015), which is consistent with the Okaro Breccia II.

Considering (1) the location of the craters excavated by phase I, (2) the confinement of the lower bed of Okaro Breccia II to the southern sector, as well as (3) the greatest deposit thickness (up to 10 m) in the southern, southwestern, and western sectors, we suggest that the phase II hydrothermal eruption began in the south. The deepest excavation of phase II is indicated by lapilli-to-large blocks derived from the Rangitaiki Ignimbrite and Rotoehu Ash, concentrated in the middle stratigraphic levels of the Okaro Breccia II in the western and southern sectors (Figs. 5 and 7; Figs. DR4 and DR5). As the phase II eruption progressed, it expanded into substrate to the north. The northern to eastern sectors of Okaro Breccia II (up to 9 m thick) contain abundant altered pumice and silicified clasts, but few large blocks (Fig. 7). This thick deposit lobe that was possibly produced by explosions directed from a crater in the northern area (Fig. 12B).

Craters from phase III excavated the ejecta from phase II (Fig. 12C). The third set of explosions excavated weak and permeable substrate, enabling efficient crater formation (cf. Montanaro et al., 2016b). Many of the recognized craters lack obvious raised rims (Figs. 10 and 11), which probably indicates breccia dispersal into lake water (Morgan et al., 2009), consistent with the sediment-covered depression and smoothed morphologies of craters from the second eruptive phase (Fig. 10). Secondary hydrothermal dissolution and collapses may also have modified the original crater shapes (Scott and Cody, 1982; Morgan et al., 2009). The thickest Okaro Breccia III occurs in the western and southwestern sectors (Fig. 7) outside the lake, and the dispersal direction was possibly affected by the presence of previous craters in the area.

Slumps and rockslide deposits at Lake Okaro provide evidence that crater enlargement continued after the eruptions. However, these deposits are confined to the discharge area of drainage networks, suggesting that mass movements occurred long after the eruptive phases. This inference is also supported by the lack of slump and landslide deposits at the steeper and deeper craters in the eastern, southeastern, and southern sectors of the lake (Figs. 10 and 11). The northern part of the lake received a sudden influx of fine sediment during the erosion of the 1886 Rotomahana Mud (a fine phreatomagmatic deposit; Hardy, 2005). On the western lake side, the sediment load from the surrounding drainage network may have destabilized semiconsolidated to consolidated Okaro breccias and older formations (e.g., Earthquake Flat Formation) on the rim of steep crater walls.

Based on breccia stratigraphy, subsurface geological structure, and new crater morphology data at Lake Okaro, we can distinguish deep “bottom-up” from subsequent “top-down” steam-driven eruptions. The history of eruptions from this area demonstrates that hazardous steam-driven explosive events may persist where a geothermal system forms in a newly disturbed site. At this locality, we see evidence for three separate episodes of explosive eruptions.

Deep-seated bedrock clasts and dominantly unaltered clasts in the first (phase I) explosion breccia show that no geothermal system was previously present at this site, but groundwater was suddenly heated and pressurized to produce a phreatic explosion. A fracture and fault network likely focused gas or magmatic fluids and produced a linear chain of craters. After this event, a new geothermal system was established within porous, weak pumice and nonwelded tuffs. After some time, during which hydrothermal alteration affected most of the Okaro area, phase II eruption was initiated from the flashing of pressured water and steam within the upper hydrothermal system. Phase II breccia contains highly altered pumice and tuff clasts, along with silicified blocks. The lower part of Okaro Breccia II has a shallow source, and higher parts have a deeper source, indicating that the locus of explosions deepened. The phase II eruption reached depths of the phase I event, as well as spreading laterally across a wide area of weak pumice deposits. Triggering of the phase II event could have involved a variety of processes, such as a seismic event or changes in the water levels of the hydrothermal system. Variably sized cluster of craters were formed during phase II. During the next pause, a widespread lake formed across the area, and hydrothermal activity continued. A third phase of explosions occurred mainly within the shallow breccia deposits of earlier eruptions. Phase III appears to have been generated by a sudden >3 m drop in the lake level.

Formation of the large crater field at Okaro was produced by steam-driven eruptions, but these eruptions had different character (bottom-up, followed by top-down) and fluids in different conditions (initially ambient and subsequently hot). This combination has been rarely reported in New Zealand or elsewhere in the world (e.g., Mary Bay crater complex in Yellowstone; Shanks et al., 2005; Morgan et al., 2009). The geological and morphological evidence from this case suggests that observable crater sizes, shapes, and distributions are partially controlled by the different excavated host rocks, as well as by the presence of pre-eruption craters and by preexisting fracture zones. This study also shows that deposit componentry can help to distinguish phreatic from hydrothermal eruptions, at least where the local stratigraphy is well known and variable enough over the depth of explosion excavation.

In a broader sense, our findings suggest that in hydrothermally active environments, the assessment of potential eruptive scenarios depends on a detailed understanding of shallow geology and the extent of the hydrothermal system. These key factors control, for instance, eruption directivity, longevity, and number of explosions, and their associated hazard. Also, once an area has been disturbed by a phreatic eruption, the subsequent formation of a new geothermal system should be very closely monitored, because later hydrothermal eruptions can be equally as large and hazardous as the initial phreatic event.

Montanaro and Cronin acknowledge funding from the New Zealand Ministry of Business, Innovation and Employment Smart Ideas grant “Stable power generation and tourism with reduced geothermal explosion hazard.” Kennedy acknowledges the New Zealand Marsden grant “Shaking magma to trigger eruptions.” We acknowledge the Bay of Plenty Regional Council, particularly Andy Bruere, and David Hamilton from Waikato University for allowing the access to the multibeam data used in this study. We also acknowledge the New Zealand Department of Conservation (Te Papa Atawhai), and Stephanie Kelly from the Rotorua Lakes Council, who supported and allowed this research to take place. We further acknowledge Larry Mastin, two anonymous reviewers, and the Associate Editor Jocelyn McPhie, whose comments significantly improved the manuscript.