Volume 124, Issue 11 p. 11380-11393
Research Article
Free Access

Seismicity and Velocity Structure of Lō'ihi Submarine Volcano and Southeastern Hawai'i

D. K. Merz

D. K. Merz

Alaska Earthquake Center, Fairbanks, AK, USA

Search for more papers by this author
J. Caplan-Auerbach

Corresponding Author

J. Caplan-Auerbach

Geology Department, Western Washington University, Bellingham, WA, USA

Correspondence to: J. Caplan-Auerbach,

[email protected]

Search for more papers by this author
C. H. Thurber

C. H. Thurber

Department of Geoscience, University of Wisconsin-Madison, Madison, WI, USA

Search for more papers by this author
First published: 16 October 2019
Citations: 5

Abstract

Hundreds of earthquakes were recorded during a nine-month ocean bottom seismometer deployment surrounding Lō'ihi submarine volcano, Hawai'i. The 12-station ocean bottom seismometer network widened the aperture of earthquake detection around the Big Island, allowing better constraints on the location of seismicity offshore Hawai'i. Although this deployment occurred during a time of volcanic quiescence for Lō'ihi, it establishes an important basis for background seismicity of the volcano. Offshore seismicity during this study was dominated by events located in the mantle fault zone at depths of 25–40 km. These events reflect rupture on preexisting faults in the lower lithosphere caused by stresses induced by volcano loading and flexure of the Pacific Plate (Pritchard et al., 2007, https://doi.org/10.1111/j.1365-246X.2006.03169.x; Wolfe et al., 2004, https://doi.org/10.1029/2003GC000618). Tomography was performed using double-difference seismic tomography and showed shallow velocities to be slower than the regional velocity model (HG50; Klein, 1981, https://pubs.geoscienceworld.org/ssa/bssa/article/71/5/1503/118231/A-linear-gradient-crustal-model-for-south-Hawaii). A broad, low-velocity anomaly was observed from 20–40-km depth, and is suggestive of the central plume conduit that supplies magma to Lō'ihi and the active volcanoes of the Big Island. A localized high-velocity body is observed 4–6-km depth beneath Lō'ihi's summit, extending 10 km to the north and south. Following Lō'ihi's active rift zones and crossing the summit, this high-velocity body is characteristic of intrusive material. Two low-velocity anomalies are observed below the oceanic crust, interpreted as melt accumulation beneath Lō'ihi and magmatic underplating beneath Hawai'i Island.

Key Points

  • An ocean bottom seismic network was deployed on Lō'ihi submarine volcano in 2010–2011
  • Lō'ihi was seismically quiescent during the nine-month deployment period
  • P wave tomography reveals low velocities beneath Lō'ihi and high velocities within its rifts and beneath the summit

Plain Language Summary

We deployed a network of ocean bottom seismometers on Lō'ihi submarine volcano for a nine-month period in 2010–2011. During that time, there were few earthquakes recorded within Lō'ihi, but many in the surrounding area. These earthquakes are thought to be associated with the bending of the Pacific Plate under the weight of Hawai'i Island rather than volcanic processes. Data from these earthquakes were used to examine the speed of seismic waves within Lō'ihi, providing clues into the volcano's internal structure. We find that Lō'ihi has many of the same internal features as the older volcanoes on Hawai'i Island. Slower velocities surrounding Lō'ihi reveal the boundary between Lō'ihi and the underlying flank of Mauna Loa volcano. Slow seismic velocities within the mantle beneath the region may be associated with the Hawaiian hot spot itself.

1 Introduction

The Island of Hawai'i is home to the most active volcanoes in the Hawaiian Islands. The island's isolated nature, combined with the lack of permanent offshore seismometers, creates a difficult environment for recording small magnitude earthquakes occurring in the offshore region. These small magnitude earthquakes are important in establishing a complete picture of the background seismic activity of the region, which defines a basis from which to compare future seismicity rates, and is essential in outlining active fault structures. In the region of Hawai'i Island, these events are useful in defining the structure of the lithosphere and its response to the stresses caused by the hot spot from below and the weight of the islands above. The Hawaiian Volcano Observatory (HVO) maintains a network of 54 seismic stations across the Big Island of Hawai'i. HVO accurately records onshore seismicity within the network, but much offshore activity goes undetected or unlocated, because of a lack of permanent offshore stations and unfavorable station geometry. The offshore detection threshold is estimated to be >ML1 (Caplan-Auerbach & Duennebier, 2001a).

Lō'ihi, the youngest Hawaiian volcano, is located ∼35 km southeast of the island of Hawai'i and lies over a kilometer below the ocean surface (Figure 1). One of the world's best studied submarine volcanoes, Lō'ihi is the youngest expression of Hawaiian volcanism, and therefore is fundamental to understanding the early stages of hot spot volcanism. Assuming that all Hawaiian volcanoes evolve in a similar fashion (Clague & Dalrymple, 1987; Moore & Clague, 1992), a proto-volcano with a Lō'ihi-type structure likely underlies all other Hawaiian volcanoes. Further, Lō'ihi formed on the submarine flank of Mauna Loa; it is unknown where Mauna Loa ends and Lō'ihi begins (Hill & Zucca, 1987), so even the volume of this seamount, and therefore its growth rate, is yet poorly constrained.

Details are in the caption following the image
Map of Hawai'i Island and Lō'ihi. Black triangles indicate stations in the HVO (onshore) and OBS (offshore) networks. Red dots are catalog epicenters for earthquakes >M2 since 1970. The black box represents the Lō'ihi region shown in Figure 2.

Much of our understanding of Hawaiian volcanoes comes from seismic studies of Kilauea and Mauna Loa, including studies of earthquake locations (e.g., Aki & Koyanagi, 1981; Got & Okubo, 2003; Klein et al., 1987; Matoza et al., 2013) and seismic tomography (Dawson et al., 1999; Lin, Shearer, et al., 2014; Okubo et al., 1997; Park et al., 2009; Thurber, 1984). From these works we know that Kilauea's south flank slips along a detachment surface at ~8–9-km depth, and that deep earthquakes occur on subhorizontal faults within the mantle (Matoza et al., 2013; Wolfe et al., 2003). From tomographic studies of Hawai'i Island we know that both Kilauea and Mauna Loa have summit magma chambers surrounded by high-velocity zones interpreted as intrusive material (Lin, Shearer, et al., 2014; Okubo et al., 1997; Thurber, 1984). Additionally, evidence exists for a smaller magma body within Kilauea's upper East Rift Zone (Haslinger et al., 2001; Lin, Amelung, et al., 2014). How early such features form in the life cycle of a Hawaiian volcano, however, is not established; imaging the internal structure of a young volcano may help address this question. Studying Lō'ihi's internal structure can help us understand when robust magma reservoirs form, the significance of mass wasting and rifting in edifice development, and shed light on the early formation of Hawaiian-type volcanoes.

We present results from a nine month deployment of 12 ocean bottom seismometers (OBSs) in an array fully surrounding Lō'ihi (Figure 2). Supplemented by the permanent onshore seismic stations of the HVO network (Figure 1), the OBS data were used to analyze offshore seismicity to establish a baseline for background seismicity at Lō'ihi. Along with high-precision relocations of earthquake locations, a 3-D inversion for the velocity structure was performed to investigate the subsurface structure of Lō'ihi, as well as fill an imaging gap between shallow (<20-km depth) 3-D tomographic studies of offshore southeastern Hawai'i (Park et al., 2007) and deep mantle (>40-km depth) tomographic studies that include the whole Island of Hawai'i (Li et al., 2000; Lin, Shearer, et al., 2014; Wolfe et al., 2009).

Details are in the caption following the image
Map of Lō'ihi with OBS locations (triangles) and station names. Dashed lines represent approximate location of Lō'ihi's rift zones. Depth scale as in Figure 1.

1.1 General Hawaiian Seismicity

Seismicity around the Big Island of Hawai'i is related to both volcanic and tectonic processes. The majority of earthquakes are concentrated around the active summits and rift zones of Mauna Loa, Kilauea, and Lō'ihi (Klein et al., 1987). Other sources include the fault zones and mobile flanks that accommodate gravitational slumping and volcanic instability (Morgan et al., 2003), and lithospheric flexure due to volcanic loading (McGovern, 2007; Pritchard et al., 2007; Wolfe et al., 2003, 2004). Flexure of the lithosphere caused by the mass of Hawai'i Island causes most seismicity deeper than 13 km to focus in zones of small, spatially distinct, reactivated faults called the mantle fault zone (Matoza et al., 2013; Pritchard et al., 2007; Wolfe et al., 2003, 2004). These deep fault zones are composed of quasi-horizontal fault planes surrounding the leading edge of the Hawaiian chain between 25- and 50-km depth. The recording and analysis of offshore activity surrounding the Big Island is limited to events near shore or events large enough to be sufficiently recorded on land-based instruments.

1.2 Lō'ihi Seismicity

The background level of seismicity for Lō'ihi is relatively low, with a few events recorded each month by the HVO network, and occasional swarms of activity (Caplan-Auerbach & Duennebier, 2001a; Garcia et al., 2005). A total of 11 earthquake swarms at Lō'ihi were recorded between 1959 and 2015, where we designate a swarm as a week with >50 located earthquakes (Caplan-Auerbach & Duennebier, 2001a). Epicenters for these swarms typically locate to the northeast and southwest of Lō'ihi's summit, despite the general north-south orientation of its two rift zones (Figure 2; Bryan & Cooper, 1995; Caplan-Auerbach & Duennebier, 2001a; Garcia et al., 2005). Due to the poor permanent station geometry relative to Lō'ihi, the hypocenters of these events have a high uncertainty in the NNW-SSE direction and depths are poorly constrained (Caplan-Auerbach & Duennebier, 2001a).

There have been three studies of Lō'ihi using OBSs. In 1986, Bryan and Cooper (1995) installed five OBSs on Lō'ihi for a period of 28 days after the onset of an earthquake swarm. Earthquakes located during this interval had hypocenters near the summit and SW flank, but the locations were not well constrained because of poor data quality, instrument failures, and the lack of an adequate velocity model for the volcano.

In 1996, following the onset of the largest earthquake swarm ever recorded at Lō'ihi, a single analog OBS was deployed on Lō'ihi's summit for two weeks (Caplan-Auerbach & Duennebier, 2001a). During the OBS deployment, 42 events were recorded on both the OBS and the HVO network. High travel time residuals on the OBS were attributed to an ill-fitting velocity model for Lō'ihi, so Caplan-Auerbach and Duennebier (2001a) developed a new velocity model by adjusting layers of the HG50 1-D velocity model (Klein, 1981). This resulted in the development of the 1-D Loihi-3 velocity model which shows decreased velocities relative to the Klein model. Locations calculated with the OBS arrivals and the Loihi-3 velocity model focus at the summit of Lō'ihi, whereas before they were oriented along the volcano's south rift (Figure 2). In addition to providing an improved 1-D velocity model for Lō'ihi, this study clearly demonstrated the benefits of offshore receivers in constraining earthquake hypocenters.

The Hawai'i Undersea Geo-Observatory (HUGO), a cabled observatory on Lō'ihi's summit, was installed in the fall of 1997 with the intention of recording real-time hydroacoustic and three-component seismic data (Caplan-Auerbach & Duennebier, 2001b). Hampered by technical difficulties, HUGO lost the use of the seismometer within a few days, but continued to record seismic events by hydrophone for the following three months. In combination with the HVO seismic network, Caplan-Auerbach and Duennebier (2001b) were able to locate 34 nearshore and offshore earthquakes. The HUGO deployment showed that the addition of an offshore receiver to the HVO network resulted in slightly different locations for events outside of the HVO array geometry, with significantly better constrained hypocenters.

1.3 Offshore Structure

Coincident with the HUGO deployment, a multichannel seismic study was conducted off the southern coast of Hawai'i aboard the R/V Maurice Ewing. Three-dimensional velocity models derived from these data show low-velocity (5.0–6.3 km/s) layers on the submarine flanks of Mauna Loa and Kilauea (Morgan et al., 2003; Park et al., 2007). These layers are interpreted as volcaniclastic sediments, and they highlight the submerged boundary between Mauna Loa and Kilauea's South Flank as an east dipping normal fault.

Data from the 1998 active source survey were also recorded on the HUGO hydrophone on Lō'ihi's summit. Caplan-Auerbach et al. (1998) used refracted arrivals to calculate 2-D velocity cross sections through Lō'ihi, showing the volcano's shallow internal structure. These profiles revealed low velocities on the shallow flanks, interpreted as mass wasting deposits. Park et al. (2007) used the same source data recorded on land-based stations to conclude that low velocities below 12-km depth beneath Lō'ihi are indicative of melt accumulation in the upper crust. Park et al. (2007) agreed with Caplan-Auerbach and Duennebier (2001a) and Garcia et al. (1998) that high velocities directly beneath Lō'ihi's summit and rift zones were consistent with intrusive materials typically found in association with Hawaiian volcanoes.

In a 2010 study, Leahy et al. (2010) used teleseismic receiver functions to image the crustal structure beneath the Hawaiian Swell. Their results suggest that magmatic underplating beneath the existing 6-km-thick oceanic crust is present below much of the Hawaiian Swell, with an approximate velocity of 7.3 km/s and an estimated thickness of 3–6 km to the south of Hawai'i Island.

Although many tomographic studies of the subaerial Hawaiian volcanoes have been performed, tomographic studies of offshore Hawai'i have only imaged above 20 km (Park et al., 2007), or below 40-km depth (Tilmann et al., 2001), and are limited in offshore resolution. The lack of offshore stations to create a wider aperture of receivers has resulted in an imaging gap between 20- and 40-km depth.

2 Data and Methods

2.1 Data Acquisition and Preprocessing

Twelve short-period OBSs were deployed on and around Lō'ihi (Figure 2) in September 2010, and all instruments were recovered in July 2011. The instruments were provided by Woods Hole Oceanographic Institution (WHOI) via the Ocean Bottom Seismometer Instrument Pool (OBSIP). The D2 series OBSs were equipped with a Geospace 3-axis short-period geophone, a hydrophone, and a six-channel Kinemetrics Q330 data logger. Data were collected at a sample rate of 100 Hz and stored internally. Due to limited battery life, most instruments stopped recording by mid-April 2011. WHOI assisted in the initial data processing, including GPS clock drift corrections and converting the recorded data into an Antelope database (Quinlan, 1998). In addition to the OBS data, HVO provided a catalog of located events with the corresponding waveforms for events recorded by the HVO seismic network during the time frame of this study.

2.2 Initial Locations

Using an Antelope database system (Quinlan, 1998), phase arrivals were handpicked on continuous data that were filtered to reduce ocean noise and identify events. Arrivals for HVO stations were revised from catalog picks provided by HVO. Once arrivals were identified, the filter was removed to ensure that P wave arrival times were not impacted by the filter. Average uncertainties for P and S wave arrivals were 0.056 and 0.28 s, respectively. Priority was given to offshore events with a minimum of four P arrivals and two S arrivals on stations in the OBS network. Where possible, data from both the OBS and HVO networks were used to best constrain earthquake locations and depths. Events with arrivals only on the HVO network and events associated with the Kilauea summit and East Rift Zone were not considered in this study.

Initial earthquake hypocenters were calculated using dbloc2, an interactive hypocenter location script included in the collection of Antelope software package programs. The location algorithm dbgenloc (Pavlis et al., 2004) was utilized in dbloc2, along with the Loihi-3 1-D velocity model (Caplan-Auerbach & Duennebier, 2001a) for offshore events, and the HG50 1-D velocity model (Klein, 1981) for onshore or nearshore events.

2.3 Cross Correlation

The GISMO suite of open-source MATLAB functions (Reyes & West, 2011) was used to cross-correlate waveforms from both OBS and HVO stations. A trigger window of 0.5 s before and 1 s after the P arrival was used with a 2–10-Hz passband filter to identify similar waveforms with a minimum correlation value of 0.6. The passband was chosen to retain the most seismic energy while filtering out vocalizations from fin whales, which are prominent in the 15–35-Hz range (Soule & Wilcock, 2013). Similar waveforms were grouped into clusters according to their similarity with other waveforms at a single receiver. Using GISMO scripts that incorporate data from an Antelope database, differential travel times were calculated for cross-correlated data and used in the double-difference relocations and double-difference tomography.

2.4 Double-Difference Relocations

Following initial locations, hypocenters were recalculated using the double-difference earthquake location algorithm hypoDD (Waldhauser & Ellsworth, 2000), which relocates neighboring earthquakes relative to one another with high precision using the difference in their travel times to a common receiver. Because this method adjusts hypocenters relative to neighboring events and not relative to the receiver, this method is not highly dependent on the input velocity model and is ideal for areas where a robust velocity model is not available. This allowed the use of the Loihi-3 velocity model for all relocations. The LSQR method (Paige & Saunders, 1982) was used to calculate relocations, as suggested by Waldhauser and Ellsworth (2000) for data sets larger than 100 events.

2.5 Double-Difference P Wave Tomography

Tomography for P wave velocity was performed using the double-difference seismic tomography algorithm tomoDD, which simultaneously relocates earthquake hypocenters and calculates a 3-D velocity model (Zhang & Thurber, 2003). This method uses differential times to derive relative locations, but also incorporates absolute arrival times to determine more accurate absolute locations relative to the receivers.

To adapt a 1-D starting velocity model to a 3-D method, a system of grid nodes was defined using the summit of Lō'ihi as the center point (x = 0, y = 0, z = 0; depth is relative to sea level), with velocities assigned to each grid node. Grid nodes in the x and y directions were spaced 5 km apart, extending 40 km in each direction from Lō'ihi's summit (Figure S1). This density ensures adequate sampling of nodes in areas of sparse ray coverage. Depth grid nodes were chosen to straddle horizontal boundaries between velocity layers of the starting velocity model in order to constrain the implied boundaries in the derived velocity model, as suggested by Evans et al. (1994). An initial hybrid 3-D velocity model was constructed using the 1-D Loihi-3 velocity model for a 30-km square area around the summit of Lō'ihi, and the 1-D HG50 (Table 1) velocity model for 35–50 km from the summit of Lō'ihi, with a lateral transition zone of one node spacing (5 km) using the average velocity between the two models.

Table 1. The 1-D Velocity Models for Lō'ihi: Lō'ihi-3 (Caplan-Auerbach & Duennebier, 2001a) and HG50 (Klein, 1981)
Depth (km) Lō'ihi −3 velocity (km/s) HG50 velocity (km/s)
0 1.5 1.9
1 2
4 4.8 6.5
4.6
11 5
15 6.9
16 8.3
16.5 8.3

To assess the resolution of the velocity model, we used the Derivative Weight Sum (DWS) as a proxy for resolution (Thurber & Eberhart-Phillips, 1999) as well as calculating checkerboard resolution models (Figures S2S4). DWS is based on the density of raypath coverage at a single velocity node, and therefore is a representation of the amount of data constraining the velocity at each node. Higher DWS typically correlates to better resolution, but low DWS may underestimate the resolution at a velocity node (Zhang & Thurber, 2007). In this study, the DWS has a maximum of 188 and a mean of 5.62 for all the sampled nodes, with the higher DWS calculated at nodes surrounded by a higher density of well-constrained events. Overall, DWS is highest between Lō'ihi's summit and the Big Island at depths between 8 and 35 km. We consider nodes with a DWS value > 10% of its mean, or 0.562, to have reliable velocities.

A synthetic database of differential times was calculated using the recorded events, differential travel times, and stations in a single iteration with the modified (±5%) checkerboard velocity model. Next, the synthetic differential travel times were used with the actual events and stations with the initial input velocity model to produce a synthetic velocity model. By subtracting the synthetic velocity model from the initial input velocity model, the ideal result will resemble the modified checkerboard velocity model. These models show good recovery near Lō'ihi at depths between 5 and 35 km, with the best recovery between Lō'ihi's summit and the Big Island (Figures S2S4). Areas of good recovery in the checkerboards also correlate to higher DWS values, as is expected (Thurber et al., 2009). No noise was added to the synthetic data, so the resolution of these models should be considered an ideal case.

3 Results

3.1 hypoDD Relocations

Relocations were initially performed using the double-difference location algorithm hypoDD (Figure 3b). Events falling in the offshore region were plentiful, with 463 events located using waveforms recorded by both the OBS and the HVO networks (Figure 3). These events included 9,675 associated P and S arrivals. When both catalog and cross-correlated times were used, we were able to relocate 281 events. For these events the arrival time RMS residuals for catalog data were reduced by 0.45 s to a final RMS of 0.31 s (variance reduced from 0.5776 s2 to 0.0961 s2), while cross-correlated RMS residuals were reduced by 0.61 s to a final RMS of 0.22 s (variance reduced from 0.6889 s2 to 0.0484 s2). Relocations support the lack of shallow (0–25-km depth) seismicity, especially directly beneath Lō'ihi's summit, as well as highlighting the tight clustering of the well-correlated clusters. The offshore seismicity recorded by the combined OBS-HVO network is dominated by deep events (∼25–40 km) that occur at a rate of ~ 2 events/day rather than in swarms, and are aligned in a subhorizontal band (Figure 4).

Details are in the caption following the image
(left) Map of Hawai'i Island showing catalog earthquakes located by HVO during the 2010–2011 OBS deployment (blue dots). (middle) Events located during this study using hypoDD and data from both the HVO and OBS networks (red dots). (right) Events located during this study using tomoDD and data from both the HVO and OBS networks (green dots). Black triangles are station locations.
Details are in the caption following the image
Final tomoDD hypocenters for earthquakes located with the OBS and HVO networks. Cross sections are shown for earthquakes within 10 km of profiles parallel to and perpendicular to Lō'ihi's rift zones. The location of Lō'ihi within the cross sections is designated “Lo,” and marked with a vertical line. Most seismicity occurs at depths > 10 km. No shallow earthquakes were recorded directly beneath Lō'ihi during the time period of the OBS deployment.

Very few seismic events were identified within Lō'ihi's volcanic edifice during the course of the OBS deployment. However, the combined OBS and HVO network located 151 events within 20 km of the Lō'ihi summit. This is more than 4 times the number of Lō'ihi-area events located by the HVO network alone (27 events; Figure 3a). Of these Lō'ihi-area events, a vast majority are located beneath the oceanic crust and are too deep to definitively correlate with Lō'ihi or its magmatic plumbing system. The events consist of mostly high-frequency waveforms with clear and impulsive P and S waves, commonly associated with crustal tectonic or volcano-tectonic seismicity. No long-period or tremor events associated with Lō'ihi were recorded. Few shallow (<20-km depth) events were recorded offshore Hawai'i (Figure 4). Additional signals detected by the OBS network but not included in the catalog include underwater landslides presumably from the periodic collapse of Kilauea's ocean entry (e.g. Caplan-Auerbach et al., 2001), whale vocalizations, and passing ships.

3.2 tomoDD Relocations and Velocity Model

To increase raypath coverage, all 463 events with initial catalog locations were used in the double-difference tomography (tomoDD; Zhang & Thurber, 2003) iterations. A trade-off analysis was used to select the smoothing parameter value, and the damping parameter was set to yield an appropriate condition number for the conjugate gradient least squares inversion. At the final iteration of tomoDD, the RMS residual for catalog data was reduced by 0.62 s to a final RMS of 0.13 s, and the RMS residual for cross-correlated data was reduced by 0.92 s to a final RMS of 0.02 s. Offshore events relocated with tomoDD generally locate within a few kilometers of the hypoDD relocations (Figures 3b and 3c). Of these offshore events, 113 lie within 20 km of Lō'ihi's summit, a substantial increase over HVO data alone.

Although both hypoDD and tomoDD provide improved perspective on offshore seismicity, the improved travel time residuals and use of the 3-D velocity model gives us more confidence in the tomoDD locations. Furthermore, the tomoDD locations cluster more tightly near the mantle fault zone identified by Wolfe et al. (2003); earthquake depths in hypoDD are significantly more scattered.

Figures 5-7 show the result of the double-difference tomography for profiles nearest Lō'ihi. The relative lack of instrumentation south of Lō'ihi's southernmost flank, as well as the lack of observed seismicity south of Lō'ihi's summit contributes to poor velocity resolution south of the summit.

Details are in the caption following the image
(a–c) East-west cross sections of Vp at y = −5 to y = 5. Lō'ihi Summit is at x = 0. Black contours are major (1 km/s) velocity boundaries and grey dashed lines are minor (0.5 km/s) velocity boundaries. The blue line indicates sea level, and the red line is the bathymetric profile. The curved line surrounding the velocity data represents areas with high resolution, defined as regions where DWS > 10% of its mean value. Areas with poor resolution are white. (d) Location of profiles a–c relative to Lō'ihi and the region offshore Hawai'i Island. Depth scale as in Figure 1.
Details are in the caption following the image
(a–c) North-south cross sections of Vp at x = −5 to x = 5. Lō'ihi Summit is at x = 0. Black contours are major (1 km/s) velocity boundaries and grey dashed lines are minor (0.5 km/s) velocity boundaries. The blue line indicates sea level, and the red line is the bathymetric profile. The curved line surrounding the velocity data represents areas with high resolution, defined as regions where DWS > 10% of its mean value. Areas with poor resolution are white. (d) Location of profiles a–c relative to Lō'ihi and the region offshore Hawai'i Island. Depth scale as in Figure 1.
Details are in the caption following the image
Cross sections of Vp at z = 4, z = 6, z = 8, z = 12, z = 16, and z = 20 km. Lō'ihi summit is at (x = 0, y = 0), indicated by the white star. Black contours are major (1 km/s) velocity boundaries and grey dashed lines are minor (0.5 km/s) velocity boundaries. The curved line surrounding the velocity data represents areas with high resolution, defined as regions where DWS > 10% of its mean value. Areas with poor resolution are white. The red line is the Hawai'i Island shoreline. High velocities suggestive of intrusive material are visible directly beneath Lō'ihi's summit, with slow velocities beneath its flanks. A low-velocity anomaly at depth extends from Lō'ihi to the north, beneath Hawai'i Island.

3.2.1 Shallow Velocity Structure (0–10 km)

A lack of shallow seismicity offshore prevents us from imaging the shallowest layers (0–4 km) in the region. The exception, however, is for Lō'ihi's summit, where increased station density provides a small window where velocities can be resolved. In this zone, a small elongate structure of high velocity (5.4–6.5 km/s) spreads north and south at z = 4 km and z = 6 km (Figures 7a and 7b). The anomaly is centered below the summit, is less than 10 km wide to the east and west, extends approximately 10 km from south to north, and is at least 2 km thick. Surrounding this anomaly, areas of relatively low velocities in the shallow layers show an increase in velocity with depth from 3.3 to ∼ 5.0 km/s (Figures 7a–7d) and are approximately 2 to 6 km thick surrounding Lō'ihi's edifice.

3.2.2 Mid-depth Velocity Structure (10–25 km)

The velocity model shows where the base of oceanic crust meets the mantle (the Mohorovičić discontinuity; or Moho) as P wave velocities increase to ∼8 km/s, consistent with upper mantle materials (Hill & Zucca, 1987). In our velocity model, the boundary is not a flat horizontal plane; high velocities begin to show at 13-km depth south of Lō'ihi and the Moho dips to 15–20-km depth closer to the Big Island (Figures 5 and 6). This is consistent with the 13-km depth defined by Park et al. (2007), the 11–14-km depth defined by Leahy et al. (2010), and similar to the 11-km depth to Moho defined by Hill and Zucca (1987). These models describe a crust that is dipping toward the island as a result of the massive weight of the islands bending the lithosphere (Li et al., 1992; Thurber & Gripp, 1988).

Within the upper mantle, two low-velocity anomalies are imaged in the mid-depth layer. One anomaly is located ~12–17 km directly beneath Lō'ihi, and consists of velocities of 6.5–7.5 km/s (Figures 5b, 8c, and 8d). The second anomaly consists of velocities of 6.0–7.5 km/s and locates below the inferred Moho, parallel to the shoreline of Hawai'i (Figure 6b).

Details are in the caption following the image
Annotated model of the velocity structure of Lō'ihi, along the NNW/SSE profile shown in Figure 5. Approximate boundaries of oceanic crust are shown with the dotted black line. Features discussed in this study (rift cumulates, melt accumulation beneath the crust, intrusive material beneath the summit, and magmatic underplating) are highlighted and labeled. The proposed plume conduit that connects Lō'ihi to the mantle plume is outlined in dashed red. Hypocenters for earthquakes located within 10 km of the profile, as calculated using tomoDD, are shown in white. Lo is Lō'ihi's summit.

3.2.3 Deep Velocity Structure (25–40 km)

Velocities below 25 km increase to 7.5–8.5 km/s, indicative of upper mantle materials (Hill & Zucca, 1987). An anomaly with reduced velocities (6.4–7.5 km/s) is observed extending from depth to the base of the oceanic crust beneath Lō'ihi and continuing northward (Figure 6).

4 Discussion

4.1 Hypocenters

The deep, horizontally aligned offshore events (Figure 4) are consistent with the deep mantle fault zones described by Wolfe et al. (2004). The lack of event density due to the relatively short observation period prevents distinct fault planes from being well defined, and instead we see a band of events occurring in unconnected patches (Figure 4). Events are frequent, occurring between 1 and 8 times per day rather than in swarms; this is typical of mantle fault zone events (Wolfe et al., 2003). A handful of events exhibit swarm-like behavior which, although not typical of mantle fault zone activity, is not unprecedented (Eaton et al., 1987).

A number of events are observed west of Lō'ihi, proximal to the western edge of the mobile Kilauea South Flank (Figure 4). Morgan et al. (2003) interpreted the western boundary of the mobile flank as an east dipping, right-lateral fault that accommodates the seaward slipping of the mobile flank as it uplifts and folds the volcaniclastic sediment of the flank into the prominent Papa'u Ridge. Although activity has been recorded in this region in the past, these events are the first to be well-located along the hypothesized boundary fault.

4.2 Velocity Model

The tomographic velocity structure presented in this paper reveals features related to Lō'ihi volcano, as well as its connection to the volcanoes of Hawai'i Island and the underlying plume. The offshore region SE of the Big Island has been the subject of several seismic studies (Hills et al., 2002; Morgan et al., 2000, 2003; Park et al., 2007, 2009), but this is the first to provide a detailed look at the velocity structure of Lō'ihi itself.

Low velocities surrounding Lō'ihi are consistent with the materials comprising Mauna Loa's submarine flank (Hill & Zucca, 1987). Although we have poor resolution near the seafloor, other studies have shown that the shallowest layers of Lō'ihi consist of unconsolidated material such as rubbly breccias (Garcia et al., 1993), and fragmented volcaniclastic debris from mass wasting events, the latter of which has been observed off the southern shore of Hawai'i (Morgan et al., 2003; Park et al., 2007). Although mass wasting and gravitational slumping events have steepened Lō'ihi's flanks (Fornari et al., 1988), the deep water pillow basalts and shallow blanket of hyaloclastic rocks (Hill & Zucca, 1987) of the submarine flanks of Mauna Loa and Kilauea, upon which Lō'ihi has developed, have likely influenced the composition, bulk volume, and shallow velocity structure of the submarine volcano.

A low-velocity anomaly is prominent beneath Lō'ihi's north rift, at a depth of 5–10 km (Figures 6b, 6c, and 8). We cautiously interpret this feature as the flank of Mauna Loa, on which Lō'ihi has formed. The thickness of this body is consistent with the interpreted thickness of Kilauea's South Flank, as calculated by Park et al. (2007, 2009). A lack of resolution at shallow depth makes it difficult to establish the extent of the anomaly north of Lō'ihi. Velocities within this feature are low (~3.5–4 km/s), suggesting that it is composed of highly fractured or fragmented material.

Although much of the shallow region around Lō'ihi is not imaged in our model, we find high velocities immediately beneath the seamount's summit, extending to the north and south beneath its rift zones (Figures 6b and 8). This shallow (4–6-km depth) high-velocity anomaly closely parallels Lō'ihi's known rift zones, with velocities similar to those observed within Kilauea's rift zone (Got et al., 2008; Haslinger et al., 2001; Lin, Amelung, et al., 2014; Lin, Shearer, et al., 2014; Okubo et al., 1997), and consistent with basaltic dikes (Planke et al., 1999). Higher velocities (6.5 km/s) are also imaged directly beneath the summit and may reflect the presence of intrusive rocks such as gabbro (Okubo et al., 1997). Intrusive material within the old ocean crust was invoked by Park et al. (2007) to explain a high-velocity (6.5–6.9 km/s) feature at 7–12-km depth beneath Lō'ihi. Alternatively, the high-velocity zone may reflect the presence of olivine cumulates at the base of a magma reservoir (Clague, 1988; Hill & Zucca, 1987; Lin, Shearer, et al., 2014; Thurber, 1984), provided that some partial melt is also present to explain the relatively lower speeds.

Two depths have been proposed for magma reservoirs at Lō'ihi. Geochemical data from the 1996 eruption suggests that there is a reservoir at 8–9-km depth (Garcia et al., 1998), placing it between our high-velocity anomaly and the inferred top of the oceanic crust. A study of Lō'ihi's summit morphology, however, suggests the presence of a shallower magma body, lying between 1 and 2.5 km beneath the summit, directly above the imaged high-velocity body (Clague et al., 2019).

We define the top of the mantle and base of the oceanic crust as the depth at which velocities exceed 8 km/s. A north-south cross section through Lō'ihi's summit (Figure 6b) suggests that this boundary deepens to the north, as the plate is depressed by the load of the Big Island. At the southern end of Lō'ihi, the mantle lies at ~15 km below sea level, or 10 km below the seafloor, consistent with the model presented by Park et al. (2007). On the northern end, however, we reach mantle velocities at depths deeper than proposed by Park et al. (2007), at ~20 km. Earlier geophysical studies proposed that the mantle is depressed to ~15–20 km beneath the Hawaiian Islands (Thurber & Gripp, 1988; Watts et al., 1985); that the crust is deepest near Hawai'i Island is consistent with its significant size.

Cross sections through Lō'ihi's summit show that mantle velocities (~8 km/s) are depressed directly below the seamount (Figures 5b and 6b), a feature also noted by Park et al. (2007). We propose that this reflects the presence of melt; such a reservoir was proposed by Clague (1988) based on the petrology of olivine cumulates in Lō'ihi lavas. Another low-velocity anomaly lies to the north, beneath the shore of Hawai'i Island (Figure 6b), and is consistent with an estimated 3–6-km-thick layer of magmatic underplating proposed by Leahy et al. (2010).

A large low-velocity anomaly extends north from ~30- to 40-km depth near Lō'ihi, shallowing near the coast of Hawai'i Island (Figures 6 and 8). This anomaly extends to the north and northwest, beneath the Pahala region of Hawai'i. The Pahala region experiences frequent tremor and deep earthquakes thought to be associated with magmatic recharge (Aki & Koyanagi, 1981; Wech & Thelen, 2015). We propose that the velocity anomaly in this region represents Lō'ihi's plumbing connection to the Hawaiian plume. Petrologic and geochemical studies of Lō'ihi lavas (Garcia et al., 1993; Kurz et al., 1983) have shown that despite the proximity of Lō'ihi to the older volcanoes on Hawai'i Island, the young seamount has relatively primitive magmas and therefore its own plumbing system. Assuming Lō'ihi has evolved in the same fashion as the other Hawaiian volcanoes, a connection between Lō'ihi's shallow magma reservoir and the hot spot at depth is expected and would consist of focused flow of melt within dikes (Rubin, 1995). Although the lack of S wave velocities in our model prevents us from investigating Vp/Vs ratios that might allow us to investigate the composition or degree of melt in this anomaly, its large size suggests it is a connected system of dikes and sills. Assuming that Lō'ihi's plumbing system has physical characteristics similar to those at Kilauea and Mauna Loa, it is likely that in the rigid lithosphere there exist melt-filled cracks connected by faults (Shaw, 1980). The quasi-horizontal mantle fault zone between 20- and 50-km depth may have influenced the path of the anomaly, giving it a horizontal preference below 25-km depth (Figure 4).

Due to the lack of permanent offshore sensors, previous tomographic studies of the active Hawaiian volcanoes have only imaged above 20 km for offshore Hawai'i (Park et al., 2007) or are limited in offshore resolution at depth (Tilmann et al., 2001), resulting in an imaging gap between 20- and 40-km depth. As our velocity model has good resolution above 35-km depth, we are confident in the existence of this anomaly.

5 Conclusions

Despite minimal volcanic seismicity recorded near Lō'ihi volcano during an OBS deployment, 463 earthquakes were located using data from the combined OBS and HVO networks; this is nearly twice as many events as located by the HVO network alone. The events make up a background seismic catalog for offshore Hawai'i, in a time of quiescence at Lō'ihi. Along with a few shallow clusters that are likely of tectonic origin, the OBSs recorded many deep events belonging to the mantle fault zones surrounding Hawai'i Island.

The tomographic models of offshore Hawai'i fill an imaging gap between 20- and 40-km depth. Our structural model for Lō'ihi (Figure 8) shows many features consistent with previous studies. A small, high-velocity anomaly in Lō'ihi's edifice is proposed to be caused by diking within the north-south trending rift zone, with the concentration of higher velocities directly beneath the summit as evidence that Lō'ihi has an established magma reservoir. Although this anomaly in our model locates at a shallower depth than a similar feature in the Park et al. (2007) model, our models agree on the location and thickness of the oceanic crust. Below the oceanic crust, two low-velocity anomalies are present, potentially representing melt accumulation beneath Lō'ihi and magmatic underplating near the Big Island. A large anomaly with reduced velocity at depth below Lō'ihi is interpreted as Lō'ihi's connection to the Hawaiian plume.

Lō'ihi shares many characteristics with the more developed Hawaiian Islands, including its well-defined rift zones, independent magmatic plumbing system, and unconsolidated flanks. Hill and Zucca (1987) proposed a model for the evolution of Kilauea emerging from the still-growing flanks of Mauna Loa. We identify a low-velocity zone consistent with fragmental materials that may represent the Mauna Loa flank underlying Lō'ihi, although it is likely that the shallow layers of Lō'ihi are intertwined with it. Unlike Kilauea, Lō'ihi at this stage of development is not buttressed by an existing volcano and both flanks are able to move symmetrically from the central rift axis. We propose that the conduit between Lō'ihi and the Hawaiian plume is influenced by the larger volcanoes, as it appears to take advantage of the existing fractures in the mantle fault zones caused by lithospheric loading.

Although this study highlights many new features of the submarine offshore region of Hawai'i, an established network of offshore seismometers would greatly improve the resolution of both earthquake hypocenters and seismic tomography. In addition to providing a complete long-term record of offshore seismicity, an established submarine network would increase the probability of recording volcanic activity within Lō'ihi and would improve our understanding of the early stages of Hawaiian volcanic growth.

Acknowledgments

We are grateful to the captain and crew of the R/V Kilo Moana during cruises KM1018 and KM1120a. The success of the OBS deployment is due to the efforts of WHOI engineers Tim Kane and Dan Kot. Figures were generated using the Generic Mapping Tools (Wessel et al., 2013). We are grateful to Ellen Syracuse and Avinash Nayak for sharing their MATLAB scripts for plotting velocity profiles. The manuscript was greatly improved by comments from two anonymous reviewers; we thank them for their time and effort. The OBS data are publicly available via the IRIS DMC under network code 9A (2010-2011). Data used in this research were provided by instruments from the Ocean Bottom Seismograph Instrument Pool (http://www.obsip.org) which is funded by the National Science Foundation. OBSIP data are archived at the IRIS Data Management Center (http://www.iris.edu). This project was supported by NSF grant OCE-0851205.