Introduction

Interactions between hot molten lava and cold seawater give rise to hydrovolcanic explosions, also known as littoral explosions in the related literature (e.g., Moore and Ault, 1965;Sansone et al., 1991;Mattox and Mangan, 1997;Haxel and Dziak, 2005;Schlindwein and Riedel, 2010;Tan et al., 2016;Le Saout et al., 2020;Tepp and Dziak, 2021). The dynamics of the interactions are complex, and our understanding is often limited because of the lack of direct observations from the volcanic eruption. Factors limiting our understanding of the interactions include remote and hazardous conditions near the explosions, few geophysical sensors, nonuniform lava-flow morphology, and variations in sound speed profiles. Nevertheless, investigations on lava-water interactions in the mid-oceanic ridge settings, captured by different instruments such as automated underwater vehicles and remote sensing, have been carried out to constrain the physical factors such as fluxes, lava volume, flow rates, types of lava flows, grain size, and the intensity of explosions (Fröhlich et al., 1993;Caplan-Auerbach et al., 2017;Le Saout et al., 2020;Dietterich et al., 2021).

Kilauea is an active volcano on the island of Hawaii where for over decades erupted lava flows have entered the ocean (Sansone et al., 1991;Mattox and Mangan, 1997). This interaction has allowed studies of different aspects of the lava-water explosions (e.g., Mattox and Mangan, 1997;Caplan-Auerbach and Duennebier, 2001;Caplan-Auerbach et al., 2001). The 2018 Kilauea Volcano eruption was the most destructive eruption on the island in the past 200 years (Klein, 1982;Neal et al., 2019;Patrick et al., 2020). In April 2018, Kilauea's Halema'uma'u crater was 200 m wide before the onset of the major eruption. Dike intrusion in the lower East Rift Zone (LERZ) triggered a major earthquake of M w 6.9 offshore followed by caldera collapse and intense magmatism (Chen et al., 2019;Neal et al., 2019;Dietterich et al., 2021). During the eruption, lava flow entered the ocean from the LERZ, leading to hydrovolcanic explosions, as well as adding 3:5 km 2 of new land to the island and 0:76 km 3 of lava deltas offshore (Neal et al., 2019;Dietterich et al., 2021;Soule et al., 2021).

According to the accounts of the eruption (Neal et al., 2019;Patrick et al., 2020), as the caldera collapsed, magma supply from the volcano kept flowing toward Kapoho Bay between mid-May to the end of June. In early July, the flow changed its course as it moved toward the Ahalanui region. This active ocean entry served as a hotspot for the lava-water explosions. To record the seismicity beneath the submarine south flank of Kilauea and the lava-water interactions near the lava ocean entry located in the southeastern flank of the volcano, 12 short-period ocean-bottom seismometers (OBSs) with a sampling rate of 200 Hz were deployed offshore on 10 July 2018, soon after the lava flows started pouring over the coastline (Wei et al., 2021). Eleven OBSs were recovered on 16 September after the eruption stopped. The OBSs had threecomponent geophone channels (EL1, EL2, and ELZ) and a hydrophone channel (EDH) that recorded the hydroacoustic signals associated with the lava-water explosions near the shoreline (Caplan-Auerbach et al., 2019;Wei et al., 2021).

This study detects and characterizes lava-water interactions in terms of hydroacoustic signals from the hydrophone channels of the OBSs to generate a catalog of lava-water explosions and to understand their relationship to lava delta deposits. Inspection of the hydrophone records showed that the four OBSs closest to the lava ocean entry (Fig. 1; KSFL, KSFK, KSFE, and KSFF) best captured the acoustic signals from the lava-water explosions, whereas those farther away from the ocean entry had a much lower signal-tonoise ratio. We therefore focused the analysis of hydroacoustic signals on the four OBSs closest to the lava ocean entry.

Hydrophone Data and Detection of Hydroacoustic Explosions

Acoustic data recorded by hydrophones are a great source for studying volcanic activities near ocean entry. In shallow water, lava-water interactions may lead to explosions that eject chunks of molten or semi-solidified lava, which are known as "lava bombs" and are a clear hazard. On 16 July 2018, one such incident happened in the early morning when a lava bomb, almost the size of a basketball, hit a sightseeing tour boat near the lava ocean entry south of the LERZ (Fig. 1) and injured 23 people (details are provided in Text S1, available in the supplemental material to this article). According to a witness account, the tour boat was within ∼250 m of the lava ocean-entry point at the time of the incident, which was a substantial excursion from its recommended route. Because the tour boat location was in close proximity to the lava ocean entry point, we assume that the lava bomb origin location was in close proximity to the tour boat's location (this is verified subsequently), and any offset is within our margin of error (∼200 m).

To estimate the lava bomb origin, we started with identifying the explosion signal in the hydrophone channels, considering the event that hit the tour boat. To identify the lava bomb signal, we used the ground truth time and location from the metadata embedded in the cell phone pictures captured by a passenger on the tour boat on her smartphone as a reference (see Text S2). Hydrovolcanic explosions can be characterized by a broad frequency range between 20 and 80 Hz and a very short duration (<∼2-3 s). Spectrograms (Fig. 2) show the hydroacoustic data within a 50 s time window surrounding the time of the lava bomb incident (as observed from phone's metadata) in the four hydrophone EDH channels. Hydrophone channel data were recorded with a sampling rate of 200 Hz. These data were bandpass filtered between 20 and 80 Hz to enhance hydrovolcanic explosion signals and reduce those generated by ground motion or seismic noise.

After manually identifying the event in multiple hydrophone channel recordings and noting their arrival times, we picked the signals at each station and used them to determine the event location (see Text S3 for details). The location was found to be within ∼190 m from the tour boat's location, confirming that we identified the correct signals for the explosion. Because of the complexity and variability of waveforms, we calculated the arrival delay times between the four station (KSFE, KSFF, KSFK, and KSFL) pair combinations by picking the "peak" values of the power envelopes (see Text S3) from their associated windowed waveforms. Because the OBSs were deployed at different depths (Wei et al., 2021), the station locations and depths were crucial to calculating the differential travel times between each station pair. The delay times between different station pairs were also calculated using the triangulation method by including both the location as well as its depth and assuming a constant sound velocity of 1.5 km/s (del Pezzo et al., 2002;Caplan-Auerbach et al., 2017;Metz et al., 2018) and compared with the values obtained from the peaks of the power envelopes. Finally, the source location was estimated by performing a grid-search evaluation (with a spacing of 200 m) along the coastline using MATLAB, with a least-square misfit to the differential travel times.

Using the determined signal characteristics, we developed a catalog of explosions for the period of OBS deployment. Automated detection of seismic or acoustic events in a noisy environment is challenging (Vaezi and Van der Baan, 2015). A commonly used method to detect seismic events is shortterm average versus long-term average (STA/LTA), a trigger algorithm that detects signals from a continuous dataset with preset values of the moving windows (Allen, 1982;Withers et al., 1998;Trnkoczy, 2012). On the hydrophone data bandpass-filtered between 20 and 80 Hz, this STA/LTA method was applied to automate the detection of lava-water explosion signals. The algorithm parameters were set via trial and error and visual inspection of the results, with 5 s as the maximum time window size (LTA) and 1 s as the short window size (STA) (see Text S4; Fig. S4). An event is labeled when the STA/LTA ratio is larger than 4. This STA/LTA detector was applied to the complete span of the two-month dataset for screening hydrovolcanic explosions for each station. Each station has thousands of local detections (Table S1). However, to locate the events and better constrain their relationship with the offshore lava deposit, we associate all the four stations' common detections that are within ∼1 s, after accounting for the propagation time to a single event in the catalog, using the least-square travel-time misfit. The event's source location uncertainties (root-mean-square error) were also calculated, keeping in mind the limitations due to traveltime misfit and picking errors and a constant sound velocity of 1.5 km/s. The root-mean-square magnitude, that is, inverse square law approximation, to estimate the magnitude at the receiver location (received level or RL rms ) of each event was also calculated in decibels from each signal considering a constant sound speed in the water column and by using the travel-time arrivals from the events to the station KSFK (Greinert and Nützel, 2004;Farcas et al. 2016;Tan et al., 2016;Crone and Bohnenstiehl, 2019). The magnitude was measured over a fixed 1 s window for the filtered frequency range around the event time.

Results

The final catalog includes the event location (point of origin near the coastline), origin time, least-square error estimations associated with the location, and calculated magnitude. Approximately 4300 events were detected and located over the active eruption phase between 11 July 2018 and 4 August 2018, when the fissure eruption stopped. Our catalog has the highest density of explosion records near the Ahalanui ocean entry with ∼1200 events (Fig. 3a), which served as the hotspot beginning from mid-July until the end of eruption (Dietterich et al., 2021;Soule et al., 2021). The spatiotemporal event distribution of the binned events (per 0.5 km distance) has the highest count of ∼145 events in a day (23 July), as shown in Figure 4a. The daily event counts show a brief quiet period after 23 July 2018, possibly because of a hiatus in the rate of lava ocean entry or background noise.

In the catalog, we also observe that the magnitude of events varies from 130.6 to 197.4 dB re 1 μPa, with the median and maximum magnitudes decreasing toward the end (Fig. 4b,c; Fig. S5). The highest magnitude explosions were mostly located near the Ahalanui ocean entry, where hotter lava volumes were deposited because of a fresh supply of magma during this eruption phase (Dietterich et al., 2021). The identified hydrovolcanic explosion signal of the lava bomb that hit the tour boat has an estimated magnitude of ∼151 dB, not the strongest, because the catalog contains many other higher magnitude events.

Discussion and Conclusions

The 2018 eruption of Kilauea led to the opening of several fissures, resulting in high volumes of lava reaching the coastline. According to Hawaiian Volcano Observatory (HVO) reports, the lava ocean entry at the beginning of eruption in May was concentrated near the Mackenzie State Recreation Area (Fig. 1). By early July, the active lava ocean entry shifted farther to the east, leading to new ocean entry to the south of Ahalanui. Because our OBS deployment was in early July (Wei et al., 2021) near the southeastern flank of Ahalanui, the instruments could monitor activity after 11 July until 16 September. Our catalog has the highest intensity of explosion records near the Ahalanui ocean entry during mid-July and a quick drop in the number of explosion events around 4 August 2018 (Fig. 4a), when the fissure eruption stopped (Neal et al., 2019;Dietterich et al., 2021). Therefore, our findings are consistent with the chronology of the eruption events as reported by HVO.

Identified lava-water explosions in our catalog also appeared to match the 2018 lava delta deposits near the coastline to the first order. The lava ocean entry near the MacKenzie State Recreation Area and most of the lava-entry activities northeast of Ahalanui near Kapoho bay (Dietterich et al., 2021;Soule et al., 2021) occurred before the OBS deployment. In the Ahalanui segment, the largest numbers of lava-water explosions correspond to thick lava deposits (≤200 m) and greatest coastline expansion toward the ocean with a higher slope (∼20°-35°) of contact as shown in Figure 3.

Hydrovolcanic explosions at the lava ocean entry have distinct hydroacoustic signatures than other acoustic signals such as landslides or bench collapses (Caplan-Auerbach et al., 2001, 2017;Tan et al., 2016;Tepp and Dziak, 2021), which are characterized by much longer durations (Figs. S2 andS6). A future study of the relationships between hydrovolcanic explosions and landslides or bench collapses may provide additional insights into the dynamics of lava-water interaction as well as better understanding of related hazards.

Our catalog provides useful insights on the continuous hydroacoustic monitoring of volcanic eruptions, which can possibly be used for hazard assessment and warning by the hazard monitoring agencies. The catalog (attached with the supplemental material) may also be used to facilitate studies of the seismic and infrasound signals of hydrovolcanic explosions (Wang et al., 2021;Thelen et al., 2022). Further studies on the temporal variation of erupted lava volume flowing downstream from the fissure to the ocean entry could provide additional support to the variations in the explosion frequency over time. Observations from hydroacoustic research may provide new avenues for studying volcanic eruptions, submarine landslides, bench collapses, and other coastal volcanic processes.

Data and Resources