Global Detection Capabilities of T Phases Using the IMS Hydrophone Network Open Access

The Comprehensive Nuclear‐Test‐Ban Treaty Organization (CTBTO) International Monitoring System (IMS) is a global network of seismic, hydroacoustic, infrasound, and radionuclide stations designed to continuously monitor the world for nuclear explosions underground, in the ocean, and in the atmosphere. Since 2001, the hydroacoustic component of the network (see Fig. 1) has been recording underwater signals propagating in the oceans using up to six stations of hydrophones suspended in the water column of the oceans and five coastal seismometer stations deployed at remote islands (Lawrence and Grenard, 1998). T phases, seismic waves that convert to underwater acoustic energy in the ocean, are a key feature among the signals recorded by the hydroacoustic component of the IMS network. T phases can originate from earthquakes or underground explosions rather than in‐water explosions (classified as H phases by CTBTO). At the CTBTO, T phase detection and classification have a pivotal role in verifying the underwater explosive nature of signals recorded at hydrophone stations.

T phases are low‐frequency acoustic waves (<40 Hz) that can travel for thousands of kilometers with low attenuation in the ocean’s Sound Fixing and Ranging (SOFAR) channel (e.g., Ewing et al., 1952; Okal, 2008). Traveling at the speed of sound in the water, ∼1.48 km/s (Munk and Forbes, 1989), T phases have a propagation speed lower than P and S seismic waves propagating in the crust. T phases can preserve high‐frequency near‐field information lost by seismic waves propagating in the solid Earth with higher attenuation (Bohnenstiehl et al., 2002). This ocean property, as a complement to the nuclear explosion verification mission of CTBTO, provides the possibility to use T phases, for instance, to estimate submarine earthquake rupture model parameters (e.g., Tolstoy and Bohnenstiehl, 2005; Yun et al., 2022), for seismic hazard assessment in coastal areas (e.g., De Caro et al., 2024) and to give a valuable contribution to tsunami warning systems (e.g., Okal et al., 2003; Carrasco et al., 2020). Indeed, in addition to monitoring the oceans for signs of nuclear explosions, IMS hydroacoustic data have been used for a broad range of civil and scientific applications such as studies on marine mammals (e.g., Pinto and Chandrayadula, 2021), shipping noise (e.g., Harris et al., 2019), submarine volcanic activity (e.g., Matoza et al., 2022), or glaciers and icebergs (e.g., Evers et al., 2013).

Recently, T phases recorded by the IMS hydrophone network have been used to infer basin‐scale deep‐ocean temperature changes from the travel times of T phases generated by repeating earthquakes (Wu et al., 2020; Peng et al., 2024), or to study long‐range 3D underwater acoustic propagation effects (Heaney et al., 2017; Oliveira et al., 2021). A significant benefit of these studies is their use of natural seismic activity, which is abundant in the underwater soundscape, rather than artificial sound generation that can harm the ocean environment (Duarte et al., 2021). In addition, these recent studies underscore the importance of improving our understanding of the T‐phase detection capabilities of the IMS hydrophone network.

The conversion of seismic energy from submarine earthquakes to hydroacoustic T phases does not always occur immediately at the earthquake epicenter (e.g., Talandier and Okal, 1979; de Groot‐Hedlin and Orcutt, 2001). Instead, this conversion process can take place hundreds of kilometers away from the epicenter, often in areas characterized by large bathymetric gradients such as seaward‐dipping slopes along subduction zones, continental shelves, islands, and seamounts (e.g., Bottero et al., 2020; Oliveira et al., 2024). The uncertainty in identifying these conversion zones can lead to errors in relating T‐phase arrival time and azimuth to the location of the signal‐generating event. In contrast, some ocean regions exhibit a more straightforward relationship, where the earthquake epicenter can be approximated as the sole location of T‐phase sources (e.g., de Groot‐Hedlin, 2020; Godin, 2021; Shen et al., 2024). Because of the complexity of the T‐phase generation process, in general, they are not used during the estimation of the location of events by automatic data processing systems or in the same refined estimation process that is part of a review by experienced hydroacoustic analysts.

This study initially analyzes T phases recorded at the six IMS hydrophone stations to assess the global network’s coverage and the detection capabilities of individual stations. Specifically, we examine T phases detected at hydrophone stations and associated with analyst‐reviewed bulletins from October 2001 to March 2024. Second, we aim to identify the earthquake events responsible for major T‐phase peaks in analyst‐reviewed bulletins. Finally, we investigate if, while T‐phase detection itself is not network‐dependent, associating them with events relies on IMS seismic network sensitivity to locate seismic events.

We begin by describing the hydroacoustic network and how the recorded data are processed in the International Data Centre (IDC), first automatically processed in near‐real time and later reviewed by human analysts. Understanding these foundational aspects is crucial for interpreting the detected T phases. Subsequently, we present our results. Finally, we summarize the key conclusions of our study.

A total of eleven IMS hydroacoustic receiver sites are in operation worldwide (Fig. 1a). The roll‐out of the network occurred progressively over almost two decades and concluded with the installation and subsequent certification of IMS station HA04 (Crozet Islands) in 2017.

Six of the hydroacoustic receiver sites are hydrophone stations (see Table A1 in the Appendix for station locations and dates of entry into operation). Except for HA01 (Cape Leeuwin), two sensor triplets—designated “N” (north) and “S” (south) as part of their IDC identifier—are located at opposite sites of remote ocean islands to avoid acoustic blockage and maximize aperture sensitivity (Fig. 1b–h). Hydrophones are bottom‐moored, floated in the local SOFAR channel axis, and organized in a quasi‐planar triangular configuration of 2 km equidistant spacing (Fig. 1i). Acoustic pressure is sampled at 250 Hz and transmitted in near‐real time to the IDC in Vienna for automatic processing and analyst review (e.g., Hanson et al., 2001; Le Bras et al., 2024).

Five so‐called T stations complete the hydroacoustic network. Typically deployed close to the shoreline of remote islands and operating at a sampling rate of up to 100 Hz, these land‐based seismometers are optimized for the detection of waterborne phases following the conversion of energy at the ocean–land interface (e.g., Okal, 2001). T stations are less effective than their hydrophone counterparts, lacking the sensitivity of a sensor directly suspended in the water column as well as the azimuthal constraints that can be derived from array‐type processing of triplet data; they are, however, more cost‐efficient to install, maintain, and repair.

The workflow of data processing at IDC consists of two main parts: automatic processing and interactive analysis by human specialists, referred to as analysts. The automatic processing component can be further divided into two phases: station processing and network processing. Following interactive analysis, an additional automatic processing phase, postlocation processing, takes place. However, for this study, we will focus on the initial automatic processing and analyst review.

The main task of station processing is to detect signals of interest in the waveforms arriving at the IDC from the IMS stations, calculate values for prespecified signal features (such as arrival azimuth, slowness, signal‐to‐noise ratio, amplitude, period, and many others), and classify these detections as phases of interest or noise phases. A definition of event, signal, detection, arrival, and phase is provided in the Appendix. Although waveforms arrive at the IDC continuously in near‐real time, station processing is performed in batch mode on time intervals of 10 min for seismic and hydroacoustic and 30 min for infrasound waveforms. The initial detection step for hydrophone waveforms is performed by the detection and features extraction algorithm using a short‐term average over long‐term average power detector (STA/LTA) with a 10 s STA window and a 150 s LTA window, deviates from the standard values of 1 and 30 s used for seismic data processing (Preparatory Commission for the Comprehensive Nuclear‐Test‐Ban Treaty Organization [CTBTO], 2002; Le Bras et al., 2024). All detections, that is, not only those classified as phases but also those classified as noise, are stored in the standard list of signal detections, a standard product of station processing.

Phases detected during station processing at different stations are grouped during network processing (Preparatory Commission for the CTBTO, 2001; Le Bras et al., 2021), the objective of which is to build and locate events that are hypothesized as having generated the grouped signals. It should be noted that because waves travel at different speeds in different media (solid Earth, oceans, and the atmosphere) until they are recorded at the IMS stations, signals due to the same event may arrive at the IMS stations and consequently at the IDC from minutes until many hours after the event time, something that significantly affects network processing. For this reason, network processing is performed multiple times during automatic processing.

The first instance of network processing takes place about one hour after the event time and includes detections made on waveforms recorded at stations of the primary seismic and hydroacoustic networks.

Events created during this first pass make up the standard event list 1 (SEL1). The second network‐processing run is performed approximately four hours after the event time. Detections used in this second run include detections at stations of the auxiliary seismic network located at distances up to 30° from a SEL1 event and at infrasound stations; detections made by auxiliary stations are not used to build new events but rather only to improve the event location accuracy. In addition, detections made in late‐arriving waveforms recorded at the primary seismic and hydroacoustic network stations are used to generate a second, more accurate event list, SEL2. Finally, a third run of network processing takes place about six hours after event time, including detection from any other late‐arriving waveforms, and generates SEL3 which is the most accurate automatically generated IDC event bulletin.

For an event to be built during the automatic processing phase of the IDC pipeline certain phase‐dependent requirements have to be met. In particular, every phase that can participate in building an event is assigned a weight (see Table 1) and for an event to be built the sum of the weights must be higher than 3.55. This means that two primary (first‐arriving) phases, such as P or PKP, detected at seismic array stations are sufficient to build an event, but at least three primary phases are required to build an event if one of the stations is a three‐component station. Similarly, two H phases (that is, hydroacoustic phases that originate from in‐water events) detected at different stations can also build an event. Phases that are not used to build events may be later used to constrain the location of previously built events or simply associated with those events, without affecting the estimate of event time and location; in the former case, they are called defining phases, whereas in the latter they are called nondefining phases. T phases are neither used to build events nor to improve the location of events built by the IDC automatic processing system, that is, they are nondefining phases. However, during interactive analysis, the analysts are allowed, following specific guidelines, to rename phases. This is further discussed in the Analyst review section.

Expert analysts at IDC review the results of automatic processing (SEL3). They combine their extensive experience with a set of instructions that may be either mandatory (rules) or recommended (guidelines) (Preparatory Commission for the CTBTO, 2012). During the interactive review, they may delete false events, split automatically built events into multiple events, or merge multiple events into one; they may build events missed by the automatic system using existing automatically detected phases or pick detections missed by the automatic detector. They may modify events by removing associated phases or associating them with other events; they may re‐pick the arrival onset time of phases, or even rename phases. When the first cycle of event review is completed, results are verified by a lead analyst, with more experience in waveform data analysis. All accepted events make up an intermediate bulletin, the Late Event Bulletin (LEB). A set of event definition criteria is applied to the LEB events. Events that consist of three or more primary stations and an event weight ≥4.6 are stored as the Reviewed Event Bulletin (REB), which is another standard IDC product. The REB is therefore a subset of the LEB and includes about 75% of the LEB events in the last ten years.

Phase renaming is a task that is undertaken with extreme caution by the analysts as it may turn a nondefining phase into a defining one and therefore render it eligible for event location. For instance, during the calendar year 2023, more than 3% of the phases associated with events that were initially classified as T phases during automatic processing were renamed by the analysts into other phases (H, P, Pn, Sn, and others). Analysts may also manually add H and T phases to the event if clear but undetected arrivals are observed in the waveforms and identified by analysts as related phases. H and T phases are identified based on signal characteristics, like the duration, spectral components, or spectral scalloping, which indicates a bubble pulse and is characteristic of an H phase. An example of a waveform and associated spectrogram of T and H phases recorded at H10N is shown in the Appendix (see Fig. A1) and more examples can be found in Le Bras et al. (2024) and Schwardt et al. (2022). Identification of H and T phases is not always straightforward as the propagation conditions such as partial blockage by islands may influence the frequency content. Analysts look at events in a holistic manner. They can assign a phase with an appropriate name based on the observations on other associated stations and check the propagation path for obstacles to the in‐water signal. Analysts can also identify seismic signals at hydroacoustic stations that may have been identified as waterborne signals by the automatic system. This can be done by considering the distance of a hydroacoustic station from the epicenter and inspecting the signal appearance. When an event’s epicenter is close to the shoreline and/or the event is large, seismic phases can be observed at several hydroacoustic stations. For instance, seismic arrivals were detected on all 11 IMS hydroacoustic stations from the $Mw$ 7.4 Antofagasta Chile earthquake (on 19 July 2024). T phases are associated with seismic events; however, as nondefining phases, they do not contribute to the event location. If the event location is based on a few seismic arrivals, an unusually large time residual of a T phase may indicate the necessity of improving seismic arrivals’ parameters. Moreover, although T phases are by default nondefining phases, in some particular cases (a total of 70 LEB events since 2001) their arrival time may be used to define and locate the event. Some of the main challenges in reviewing T phases detected in the automatic processing include picking the arrival time. A large section of the waveform record may be identified as a single T phase but it may include sound arriving from a variety of paths. The earliest portions of the signal may have traveled via high‐speed crustal paths before coupling into the water at distant locations such as seamounts closer to the recording station than the epicenter. Such paths result in arrival at the recording station earlier than sound which coupled into the water close to the epicenter and then traveled at the lower waterborne speeds. Later portions of the waveform may include sound that has traveled outside the geodesic path connecting the event to the station, and then been reflected off seamounts and coastlines before traveling to the recording station. In addition, sound may travel along waterborne paths from the event to the recording station, but not follow the geodesic path, that is, being diverted by horizontal sound‐speed gradients caused by spatially varying sea–water temperature and salinity.

The coverage of IMS hydrophone stations significantly depends on the station’s location. T‐phase detections by hydrophone stations associated with LEB events are depicted in Figures 2 and 3 with a 0.5° grid resolution. These figures illustrate the global (first panel, “All”) and station‐specific coverage of these events, which are mainly related to earthquakes concentrated along tectonic plate boundaries, particularly midocean ridges and transform faults, where most submarine earthquakes occur. Looking at station‐specific coverage, H01W mostly records T phases from the Indian Ocean, south of Australia, the Pacific Antarctic Ridge, the East Pacific Rise, and the South Atlantic part of the Mid‐Atlantic Ridge. Long T‐phase propagation paths to H01W have been recorded from the Pacific coast of Central America (∼16,500 km path) and the Mid‐Atlantic Ridge in the North Atlantic (∼17,500 km path). Hydrophones from H03N, H03S, H11N, and H11S mostly receive T phases from the Pacific Ocean, which includes the active seismic areas of the ring of fire. On the other hand, H04N, H04S, H08N, and H08S hydrophones have been mostly recording T phases from the Indian Ocean, south of Australia, and the Atlantic–Indian Ridge. Worth noting are T phases in the LEB recorded at H08N and H08S from some few earthquakes in the Pacific coast of Mexico, leading to a propagation path of ∼21,300 km, exceeding the antipodal range. In the South Atlantic Ocean, H10N and H10S hydrophones record T phases from the north and south Atlantic Ocean. Notably, these hydrophones also record a high number of T phases from the very active Kermadec Arc region in the southwestern Pacific Ocean, located ∼15,000 km away.

The occurrence of large earthquakes is often followed by numerous subsequent aftershocks, increasing the number of events and associated phases in the LEB and consequently increasing analysts’ workload. Figure 4 investigates the relationship between major peaks in seismic and T phases in LEB with major earthquake events that occurred between 1 October 2001 and 1 March 2024. The analysis started by focusing on earthquakes with $Mw≥7.5$ and depths less than 100 km, according to the U.S. Geological Survey National Earthquake Information Center earthquake catalog (U.S. Geological Survey [USGS], 2024). Figure 4a shows the total number of seismic phases (excluding surface waves) per month in LEB from IMS seismic primary stations and the number of active IMS seismic stations (primary and auxiliary stations). Similarly, Figure 4b shows the number of T phases per month in LEB from the hydrophone stations and the number of active hydrophones. The major peaks in seismic phases in LEB correlate well with the occurrence of $Mw≥7.5$ earthquakes. However, there is no clear correlation between peaks of T phases and major earthquakes before 2008, likely due to the limited availability of hydrophone stations during this period; however, two notable exceptions are observed: the $Mw$ 9.1 Sumatra Indonesia mega earthquake on 26 December 2004, and the $Mw$ 8.6 Sumatra region earthquake on 28 March 2005. In contrast, a strong correlation exists between major peaks of T phase and $Mw≥7.9$ earthquakes after 2008. Moreover, after the 2011 Tohoku earthquake, one very dominant T‐phase peak per year can be observed from 2012 to 2016, and these coincided with submarine earthquakes of $Mw≥8.0$⁠.

Although most T‐phase peaks correlate well with $Mw≥7.5$ earthquakes, few peaks are not well explained under this rule leading us to also analyze $7.0≤Mw≤7.4$ earthquakes. A T‐phase peak in September 2016 correlates well with a group of submarine earthquakes (group 2 in Fig. 4) that occurred from the middle of August to 1 September 2016, which includes the earthquakes in Loyalty Islands (⁠$Mw$ 7.2), South Georgia Island (⁠$Mw$ 7.4), North Ascension Island (⁠$Mw$ 7.1), and northeast of Gisborne New Zealand (⁠$Mw$ 7.0). The entering into operations of HA11 in December 2007 and the subsequent occurrence of $7.0≤Mw≤7.4$ submarine earthquakes from February to July 2008 led to a steep increase in T phase associated in LEB (group 1 in Fig. 4).

Overall, the largest peak observed in monthly seismic phases and T phases occurred in March 2011, linked to the $Mw$ 9.1 Great Tohoku series of earthquakes. This is explained by the fact that the aftershocks of the 2011 Tohoku earthquake were more active and energetic than those of the 2004 Sumatra and 2010 Chile earthquakes (Hirose et al., 2011). Between 1 October 2001 and 1 March 2024, a total of 1,003,543 T phases recorded by the IMS hydrophones were associated in LEB, 13,942 of which in March 2011, with a daily maximum of 1,589 T phases on 26 December 2004 linked to the $Mw$ 9.1 Sumatra Indonesia mega earthquake.

Figure 5a shows that the number of events detected on the entire network grew over the period of study. The dots in the figure show daily numbers of LEB events, with a 91‐day running median plotted as a line running through those dots. Initially, around 50 events were recorded each day by the network but this number climbed to around 140 events per day by the beginning of 2024. This increase is a consequence of the increasing number of stations in the IMS network improving detection capabilities, especially for smaller magnitude events (Qorbani et al., 2024). The lower curve in Figure 5a shows the number of events recorded on the network that have at least one T phase associated with them in LEB. This initially increased but flattened out after 2008, which matches the time at which HA11 came online. H11N and H11S are located at a site with unblocked paths to most of the seismically active regions in the Pacific Ocean (see Fig. 3) and consequently ideally situated for the detection of T phases. This is also shown in Figure 5c, which is a histogram of the number of T phases detected and associated with events, shown as a function of detecting station and time. Both hydrophone triplets at HA11 are consistently shown to detect a larger number of T phases than any other hydrophone station triplet. Figure 5b shows the numbers of events, normalized to unit maximum, and the proportion of all events that have a T phase associated with them. This proportion is shown to stabilize at around 0.2 after the time at which HA11 became operational.

The curve of the total number of events, shown in Figure 5a illustrates an interesting feature in that there is an oscillation superimposed on the longer‐term, steady rise in the number of events. This is only visible in the 91‐day averaged curves and indicates that the number of events detected on the IMS network changes cyclically with a yearly period. Further investigation of this effect revealed several seismic stations with yearly cycles in the number of seismic phases detected and subsequently associated with events by network processing. Examples are shown in Figure 6 where four seismic stations (see Fig. 1 for stations location) exhibiting cyclic association are highlighted. The seismic station YKA (Yellowknife, Canada) shows a very strongly periodic signal, with the number of associated phases per day varying from minimum values of around 20 in the northern‐hemisphere summer to maximum values of more than 50 in the winter. Similar behavior is observed for seismic station MKAR (Makanchi, Kazakhstan). An equally clear, but smaller‐amplitude, oscillation is shown in the number of associated phases at the MAW seismic station (Mawson, Antarctica) but in this case, maximum values occur in the northern‐hemisphere summer and minima in the northern‐hemisphere winter: in antiphase with the two northern‐hemisphere stations.

This periodic behavior is interpreted as indicating seasonality, not in global seismicity, but in the IMS network’s ability to detect that seismicity. Stations YKA, MKAR, and MAW share an important aspect of their locations in that they are all close to large bodies of water: The Great Slave Lake, Lake Alakol, and The Southern Ocean, respectively. Periods of high numbers of associated phases at these stations match periods in which the adjacent bodies of water that are the predominant noise source are frozen (Bahavar and North, 2002; Koper et al., 2009). Periods of low association are associated with the presence of liquid water and consequently with local wave‐generated noise. The oscillations observed in the total number of detected events are a consequence of the fact that stations like YKA and MKAR are “backbones” of the IMS seismic network, that is, without them, many small events would not be detected on enough IMS stations for acceptable event solutions to be formed (Qorbani et al., 2024). The cyclic detection capabilities of these “backbone” stations, which are in the northern hemisphere, impose a similar, winter‐peaking pattern on the total number of phases detected on the network.

This effect results in cyclic detection performance being observed on stations that are not subject to seasonal variations in local noise. An example of this is shown in the curve for seismic station ASAR (Alice Springs, Australia) in Figure 6. This station is in a hot, arid region, far from any mass of water, but Figure 6 shows peaks in associated phases that correlate with the peaks at YKA and MKAR. This behavior arises because, in periods where noise is low at those stations, ASAR detects signals from small events that can be resolved by the IMS network, using signals from YKA and MKAR. Similar seismic events occurring in the northern‐hemisphere summer may result in detectable signals at ASAR, but the absences of detections at stations such as YKA and MKAR result in the events going undetected and a subsequent reduction in the number of associated phases at ASAR.

The yearly period in the number of associated phases at ASAR is shown perhaps more clearly in the lower‐right panel of Figure 7. This panel shows the spectrum produced by Fourier transforming a time history of the number of events. The spectrum is plotted, not as a function of frequency but as a function of period, measured in days. The dashed lines show periods associated with yearly cycles (365 days) and lunar cycles (27.3 days). The bottom‐right panel shows results for ASAR and a peak in the spectrum is visible around the yearly period. Other panels in Figure 7 show spectra for the number of phases associated at hydrophone stations and, in some cases, yearly peaks are also observed: H03S, H11N, and H11S are particularly good examples. This can be interpreted as arising from a similar effect to that observed at ASAR. That is, the periodicity arises because of cycles in network detection capability, not because of cycles in local noise conditions. The strength of the peak varies from station to station, it is proposed, because of different stations’ abilities to resolve events for which presence in LEB is more or less strongly affected by noise conditions at stations such as YKA and MKAR.

Other peaks exist in the data shown in Figure 7 and the locations and shapes of these peaks vary from station to station. Further studies would be required to propose underlying physical phenomena in each case. A particularly interesting, and as‐yet unexplained, feature of the same data is the repeated presence of a peak at a period of ∼520 days (shown by a dashed line labeled 520), particularly strong at HA01, HA08, H10N, and HA11.

This study analyzed T phases detected at IMS hydrophone stations that were associated with LEB events from October 2001 to March 2024. The analysis aimed to evaluate the T‐phase coverage of the IMS hydrophones network. Overall, global ocean coverage is provided by combining the six IMS hydrophone stations, where T phases are mainly related to earthquakes concentrated along tectonic plate boundaries, particularly midocean ridges and transform faults, where most submarine earthquakes occur. However, T‐phase coverage by single stations significantly depends on the station’s location, with in some cases T phases being recorded after propagating in the ocean for more than 20,000 km, exceeding the antipodal range.

Although the major peaks in seismic phases in LEB correlate well with the occurrence of $Mw≥7.5$ earthquakes, there is no clear correlation between peaks of T phases and major earthquakes before 2008, likely due to the limited availability of hydrophone stations during this period; however, two notable exceptions are observed: the $Mw$ 9.1 Sumatra Indonesia mega earthquake on 26 December 2004, and the $Mw$ 8.6 Sumatra region earthquake on 28 March 2005. In contrast, a strong correlation exists between major peaks of T phases in LEB and $Mw≥7.9$ earthquakes after 2008. Overall, the largest peak observed in seismic phases and T phases corresponds to the 2011 Great Tohoku series of earthquakes.

Although the detection of T phases is independent of the seismic network sensitivity, the association of T phases to events depends on the ability of the IMS seismic network to build events. This is because except for a few exceptions, events are not located based on arrival times and back azimuths of detected T phases. For instance, although HA11 does not present seasonality in automatic T‐phase detection, the association of these T phases to LEB events is sensitive to the contribution of the seismic array station Yellowknife to build events in the North Pacific Ocean. This could be because, during the North Hemisphere Winter period, the Great Slave Lake is frozen, turning Yellowknife station less noisy and consequently helping to build more seismic events in the North Pacific, which trigger T phases detected by HA11 hydrophones.

Traditional land‐based seismic networks and ocean‐bottom seismometers face inherent challenges in monitoring seismicity across the vast expanse of the oceans. Seismic waves attenuate as they travel through the Earth’s crust, resulting in weaker signals from distant oceanic earthquakes at land‐based stations and limiting the ability to detect and accurately locate smaller‐magnitude events occurring in remote oceanic regions. The IMS hydroacoustic network can offer a transformative solution for addressing these limitations as hydrophones deployed in the SOFAR channel leverage the efficient long‐range propagation of T phases. The 1,003,543 T phase recorded by the IMS hydrophones and associated in LEB from 2001 to 2024 provide a unique and invaluable dataset for comprehensive seismicity monitoring and improve the understanding of tectonic processes in underwater environments. Moreover, this T‐phase dataset can significantly contribute to the physical observation and modeling in the marine environment helping understand processes such as T‐phase generation and long‐range propagation and basin‐scale deep‐ocean temperature changes.

The International Monitoring System (IMS) data used in this article are available to the International Data Centre (IDC) authorized users or via the virtual Data Exploitation Centre (vDEC) platform at the Comprehensive Nuclear‐Test‐Ban Treaty Organization (CTBTO). More information on accessing IMS data via the vDEC platform is available at https://www.ctbto.org/specials/vdec/ (last accessed February 2025).

The authors acknowledge that there are no conflicts of interest recorded.

The views expressed in this article are those of the authors and do not necessarily reflect those of the Preparatory Commission for the Comprehensive Nuclear‐Test‐Ban Treaty Organization (CTBTO). The use of particular designations of countries or territories does not imply any judgment by the Commission as to the legal status of such countries or territories, their authorities and institutions, or the delimitation of their boundaries.

Definition of event, signal, detection, arrival, and phase.

• •

  Event: a physical occurrence that generates seismoacoustic energy.

• •

  Signal: a coherent perturbation of the waveform trace reflecting seismic or acoustic waves generated by events.

• •

  Detection: a seismic or acoustic signal or noise that is found in the data stream by automatic processing that meets preset criteria for signal identification and measurement.

• •

  Arrival: a detected signal that has been associated to an event.

• •

  Phase: a detected signal that is identified on the basis of its path through the Earth.