4.1 Structure of the caldera on Volcano #1
A volcano with a caldera at the summit consists of various morphologic elements including: topographic rim, inner caldera wall, caldera-bounding faults, structural caldera floor, and intracaldera fill (mainly landslide debris from caldera walls and ponded ash-flow tuff) (e.g. Lipman, 1997). The depth-converted MCS profiles demonstrate the overall structure of the caldera and intracaldera fill of Volcano #1 (Figs. 3 to 5). The topographic rim is depicted on the MCS profiles as a pointed escarpment that bounds the subsided area. At the southwestern flank, the topographic rim is buried by erupted material (Fig. 5). The V1P1 cone is composed of layered deposits; however, seismic reflection signals are highly disrupted below the V1P3 cone, making it difficult to recognize layering.
On the basis of external and internal seismic facies, four key seismic units that constitute the V1P1 cone were recognized as the caldera infill. They are seismic sequences bounded by reflecting horizons that are consistently traced in seismic profiles. These seismic units are referred to as U-1 to U-4 from oldest to youngest. Hekinian et al. (2008) suggested that post-caldera cones were constructed by occasional eruptions dominantly of coarse pyroclast-ash flow. Therefore, we interpret that seismic units of U-1 to U-4 consist dominantly of these materials. No age can be assigned to the seismic units. Hekinian et al. (2008) identified about 20 volcaniclastic units along the caldera wall and suggested a short-time interval of about decades between eruptions, based on the presence of fresh scoria and fresh lava flows and the absence of non-volcanic sediments within volcanic deposits. Therefore, they estimated the overall caldera wall to have been constructed probably during the last 200 years. The seismic units are likely to be younger than the caldera wall because they constitute the post-caldera cone of V1P1. Landslide debris is created by sliding and slumping triggered by gravitational instability during and after caldera collapse. Consequently, landslide debris tends to accumulate as debris fans adjacent to margins of the subsided block (Lipman, 1997) that would appear as a wedge thickening downwards on seismic profiles. The reflectors dipping inward below the structural rim are interpreted to be oblique sides of the landslide debris wedges (Figs. 3–5). The base of the caldera fill corresponding to the acoustic basement is not clearly distinguished. Instead, it is only locally identifiable under the V1P1 cone in the center of the caldera due to limited penetration. It is not imaged adjacent to the caldera rim where a thick accumulation of landslide debris occurs. The visible acoustic basement dips down toward the northwestern caldera rim.
Although the MCS profiles enable us to recognize the layering of the caldera infill, the caldera-bounding faults are not properly imaged, which we ascribe to the following reasons. First, the caldera-bounding faults, in most cases, are very steeply dipping or near-vertical (Cole et al., 2005). Surface seismic data do not contain energy reflected from those faults regardless of acquisition geometry. As the second explanation, the landslide debris, stacked chaotically down the caldera-bounding fault plane, scatters seismic waves. In addition, the caldera boundary consists of laterally-varying or discontinuous elements such as topographic rim, inner caldera wall, caldera-bounding faults, and slided/slumped material from the wall. Diffractions and multiple reflections from these morphologically complicated elements are difficult to remove completely and mask internal structure. We estimated the locations of the caldera-bounding faults by correlating the MCS profiles with the magnetic anomaly profiles (discussed later in Section 4.2).
U-1 is the deepest unit underneath the V1P1 cone. The detailed internal layering of U-1 is difficult to interpret due to deep burial and deterioration of data quality. It is thick in the northwestern part of the caldera and thins southeastward. The seismic facies of U-1 is characterized by multiple layers of short, complex, and irregular, high-amplitude reflections with poor continuity. The high-amplitude reflections are suggestive of highly-reflective volcanic lava sill/flow. It thus appears that U-1 consists of volcanic ash and pyroclastic sediments interbedded with volcanic and lava sill/flow. Because U-1 merges with landslide debris toward the caldera wall, we interpret that U-1 was deposited concurrently with slide debris. The interval velocity of U-1, ranging from 2.9 to 3.1 km/s, indicates significant welding and consolidation of volcaniclastics. The chaotic internal seismic facies of slide debris indicates massive and quick deposition.
U-2 and U-3 consist of low-to-high or variable amplitude, moderate-to-poor continuity reflections that have in places a wave-like character suggestive of deposition by flow processes. These units are accumulated in the entire caldera area. However, they exhibit variations in thickness. U-2 is subdivided into upper and lower layers. Overall, U-2 and U-3 form mounds on the sections, marking the location of the thickest accumulation. When an eruption occurs underwater, the ballistic dispersal of clastic materials is initially very restricted because of the enclosing water (Milia et al., 2000). Additionally, the increasing hydrostatic pressure of the water column with increasing water depth in subaqueous environments limits the ability of superheated volatiles to expand instantaneously against the ambient pressure (Cas, 1992). Thus, the post-caldera eruptive center is likely to be the site of greater thickness of a volcanic sedimentary sequence than elsewhere in the submarine caldera. The internal reflections of U-2 and U-3 display, in places, toplap truncation against the upper surface and downlap onto the underlying surface (Fig. 3), indicating that erupted material was deposited in a prograding manner away from the source.
U-4 is the uppermost unit of the V1P1 cone. The thickness of U-4, ranging from 200 to 300 m in the caldera area, does not vary significantly. This unit consists of a well-stratified succession characterized by good continuity, although a few strong amplitude reflections with poor continuity are visible. The upper part of U-4 mostly comprises weak amplitude (i.e., lack of reflectivity) reflectors that are parallel to the seafloor. The interval velocity of U-4 is around 1.5 km/s. The low amplitude and interval velocity suggest that U-4 consists of relatively homogeneous and un-consolidated volcanic ash and flows that are diffusely bedded. The nearly transparent internal seismic facies may be a result of rapid emplacement. U-4 buries the caldera wall to the west and reaches the lower part of the outer slope of Volcano #1 (Fig. 5).
The V1P3 cone is characterized by chaotic seismic facies that completely masks internal structure (Fig. 3). The uppermost part of the V1P3 cone shows weak layering with subdued internal reflections. Therefore, we anticipate that the V1P3 cone resulted from very rapid deposition of coarse pyroclastic material and ash. The chaotic seismic facies and lack of internal reflections indicate that the V1P3 cone was built by less than a few occasional highly-explosive eruptions.
4.2 Evolution of the caldera and hydrothermal alteration
Experimental studies, observations, and geophysical analysis (Cole et al., 2005; Acocella, 2007, and references therein) suggest that caldera collapse occurs on the steeply outward dipping reverse fault; a second set of outer normal faults dipping inward, associated with late-stage peripheral extension, develops at the caldera margin. Figure 6 is a conceptual model showing the evolution processes of the Volcano #1 caldera from the initiation of collapse to the present that eventually pertains to the geologic structure shown in Fig. 3. The basement floor deepening flatly toward the northwestern rim creates an asymmetric subsidence structure, which suggests that the collapse initiated in a trapdoor fashion (Fig. 6(a)) with a hinge on the opposite southeastern caldera rim. Asymmetric collapse is a common style of subsidence, either as a trapdoor along a single hinge fault or as a series of blocks (Stix et al., 2003). The zone between these inner reverse and outer normal faults became subjected to tilting and fracturing that facilitated landslides. We envisage, therefore, that the Volcano #1 caldera collapsed along the steeply-dipping reverse fault; afterwards, the overhang reverse fault scarp decayed rapidly, emplacing the thick wedge-shaped landslide debris inward of the fault plane.
While the caldera collapsed, it was filled rapidly with erupted material, volcanic sill/flow, and landslide debris deposits (Fig. 6(b)). The shape of a caldera is affected by a regional tectonic control, pre-existing structures, and the depth to which the magma body has intruded (Kusumoto and Takemura, 2005; Acocella, 2007). Volcano #1, like its caldera, is elongated to the northwest. The calderas of other volcanoes adjacent to Volcano #1 on the Tonga Arc are circular or differ from Volcano #1 in elongation direction, sensu stricto (Fig. 7). This observation may indicate that tectonic control here did not play a primary role in shaping the caldera. We further anticipate that the most deeply-subsided zone along the inner reverse fault in the northwestern part of the Volcano #1 caldera created the largest overhang which, in turn, induced enhanced gravitational instability. In this case, the caldera wall probably retreated outward more extensively to the northwest from the rim of the initial collapse as a result of sliding and slumping, facilitating the creation of the caldera elongated to the northwest. While eruptions recurred, the subsiding caldera was successively filled with erupted sequences (Fig. 6(c)) to form the V1P1 cone. Seismic sections indicate that the internal structure of the caldera infill and sequence boundaries experienced little deformation. This feature suggests further subsidence of a coherent block of rock into an evacuating magma chamber along a ring fault that induced piston/plate collapse as eruptions progressed. The undeformed unit boundaries also suggest that subsidence by sediment loading has not occurred noticeably after deposition. The experiment by Kennedy et al. (2004) indicates that highly-asymmetrical subsidence is a result of tilting of the magma chamber and the deepest point of an elongate trapdoor caldera occurs where the magma chamber is deepest. Roche et al. (2000) suggests from their experiments that shallow magma chambers with large diameters lead to coherent single-block collapse structures, while deep chambers with small diameters lead to a series of multiple nested blocks. It is possible that the collapse of Volcano #1 was associated with a shallow tilted magma chamber. While repeated eruptions occurred, a steeply-dipping fault was created in the hinged segment to complete the caldera-bounding ring-fault system to induce coherent subsidence of the caldera area. After the V1P1 cone had formed, eruptions took place along the arcuate segments of caldera margins to form a smaller cone of V1P3 (Fig. 6(d)).
Calderas are commonly the sites of geothermal activity and mineralization (Cole et al., 2005). The principal caldera fault system consisting of inner reverse and outer normal faults is the prime locus of hydrothermal fluid circulation between the surface and the magma chamber underneath, whereby cold sea water is drawn down the normal faults and returns as metal-rich hot fluid along the reverse faults, favoring the formation of volcanogenic massive sulfide (VMS) deposits at caldera wall margins (Mueller et al., 2009). A comparison of magnetization values and geologic structure seems to indicate a primary association between them (Figs. 3–5). Magnetization ranges from 0 to 8 A/m. Noticeably, well-defined magnetization lows occur along the structural rim which represents the inner ring fault. Magnetization lows are likely to be present along the entire structural rim and mark the collapse boundary. Emplacement of sulfide in the Volcano #1 caldera has been reported from submersible dives (e.g., Stoffers et al., 2006). Hydrothermal alteration of footwall rocks beneath massive sulfide deposits may lead to the destruction of magnetic phases and results in anomalously low magnetic signals (Morgan, 2012). Magnetization lows above the rim of Volcano #1, therefore, is interpreted to indicate hydrothermal activity along the principal ring fault. Magnetic lows are not observed in the center of the caldera, suggesting that the zone of hydrothermal activity is limitedly present close to the inner ring fault. Topographic displacements that appear to be the surface expressions of steeply-dipping faults are recognized on the outer slope of the volcano as well as on the inner structural rim. The coincidence of their locations with magnetization lows (Figs. 3 to 5) suggests that they constitute the outer ring fault system and are involved in hydrothermal fluid circulation that enables VMS deposits to develop.
The subsidence depth is estimated from seismic profiles by measuring the vertical distance from the top of the topographic rim to the acoustic basement. The maximally recognized depth of the acoustic basement is 2 km inward of the northwestern rim (Fig. 3). The acoustic basement may get deeper toward the caldera rim that marks the site of the principal ring fault. The ratio of the diameter (4.5~7 km) to subsidence (2 km) of the caldera ranges from 2.3 to 3.5 allowing it to be quantified as a stage 4 caldera according to Acocella (2007). A stage 4 caldera is usually associated with the largest erupted volumes, enabling the emplacement of thick accumulation of the caldera fill as observed on seismic profiles.