Coherence‐Based Characterization of a Long‐Period Monochromatic Seismic Signal

Abstract

Continuous seismic data analysis identifies signals related to physical processes within the Earth or on its surface. Characterizing seismic signals yields insights into source processes and Earth's structural features. Global seismic network analysis of long‐period (25–100 s) surface waves has detected seismic events not identified through high‐frequency body wave analysis. However, detecting long‐lasting monochromatic signals with narrow spectral peaks, which carry valuable information about geological and environmental processes, remains challenging on a global scale. We developed a coherence‐based approach to characterize long‐period monochromatic signals on a global scale. In addition to signals originating from the Gulf of Guinea, Vanuatu islands, and a submarine volcano, we observed a previously unidentified signal originating from the Canadian Arctic, likely associated with glacier dynamics. Our approach explores long‐period monochromatic seismic signals in continuous seismic data, providing a foundation for future studies to characterize the physical processes generating these signals on Earth's surface.

Introduction

Observational seismology relies on the analysis of seismic signals to characterize various physical processes occurring within the Earth and on its surface. A primary focus of seismologists is the study of earthquakes, which are cataloged in earthquake databases constructed by identifying discrete seismic phases used for event location. These data are fundamental for constraining the seismic source process, characterizing fault zone structures, and imaging the Earth's interior (e.g., Thurber, 1983; Waldhauser & Ellsworth, 2000). Traditionally, the manual picking of high-frequency body wave arrivals has played a crucial role in building earthquake catalogs. Algorithms locate seismic events by solving inverse problems based on assumed velocity models (e.g., Lin & Shearer, 2006; Thurber, 1985). Recent advances in phase-picking techniques, including matched filtering (Beaucé et al., 2018; Gibbons & Ringdal, 2006) and machine learning (Bergen et al., 2019; Kong et al., 2019), have enabled the detection of small, previously unlisted earthquakes without the need for precise velocity models. Matched filtering, which uses pre-identified seismic signals as templates, has revealed more than 10 times the number of events in original catalogs (Beaucé et al., 2024). Deep learning methods have also improved event detection by successfully training models to identify and pick P- and S-wave onsets. These approaches have provided valuable insights into the spatiotemporal evolution of earthquakes in volcanic regions (Wilding et al., 2023; Yukutake et al., 2023) and fault zones (Essing & Poli, 2022). However, both matched filtering and machine learning techniques rely on extensive libraries of event templates (e.g., McGuire, 2008; Meng & Peng, 2014) or large training datasets (e.g., Majstorović et al., 2021; Mousavi et al., 2020), which limits their ability to detect previously unknown seismic signals.

Alternative methods have been developed to detect signals lacking clear impulsive arrivals at high frequency ($${ >} 1$$ Hz), which inhibit the precise picking of seismic phases. At periods longer than 25 s, back-propagation of surface waves recorded by global seismic networks has revealed previously unidentified seismic events that are absent from traditional earthquake catalogs (Chen et al., 2011; Ekström, 2006; Poli, 2024; Shearer, 1994). Shearer (1994) detected 32 unidentified events by stacking long-period surface waves using a matched filtering technique, with most of the events located along oceanic spreading ridges in the Southern Hemisphere. Ekström (2006) analyzed global seismic data over periods of 35–150 s for 10 years, detecting about 1,300 unidentified events, primarily generated by glacial calving in Greenland and the Antarctic. Tsai and Ekström (2007) provided detailed insights into glacial dynamics using a centroid single-force model for these seismic signals. More recently, Poli (2024) applied a migration-based algorithm to long-period (25–100 s) seismic data recorded by a global seismic network, including stations in the Antarctic and Arctic, from 2010 to 2022. This analysis discovered approximately 1,766 previously unidentified events, most of which were located in polar regions—areas often missed in traditional earthquake catalogs. The events also occurred along mid-ocean ridges and tectonic boundaries, which could not be linked to known tectonic structures. Since global surface wave analyses detect signals using the shift-and-stack method, inaccuracies in velocity models can introduce uncertainties in event detection. Regional and continental-scale networks have also detected previously unidentified long-period signals (Fan et al., 2018, 2020). For example, Fan et al. (2020) divided the USArray into sub-arrays to detect signals related to seafloor landslides without requiring a known velocity model.

The aforementioned methods, however, have limited capability in detecting long-lasting and often monochromatic signals generated by volcanic (Cesca et al., 2020; Chouet, 1986; Kumagai & Chouet, 2000), environmental (Svennevig et al., 2024), and oceanic processes (Bruland & Hadziioannou, 2023; Oliver, 1962; Shapiro et al., 2006). Some of these monochromatic seismic signals are observed at periods longer than 50 s (Svennevig et al., 2024). In volcanic regions, long-period signals are frequently observed in relation to magmatic and hydrothermal activities (Chouet, 2003; Yamamoto et al., 1999). Harmonic oscillations of long-period signals—interpreted as the resonance of a fluid-filled resonator in response to a localized excitation (Chouet, 1986; Kumagai & Chouet, 2000)—offer valuable insights into magma storage and migration beneath volcanoes. These signals can sometimes be observed globally, as demonstrated by the Mayotte, where long-period tremors with a 16 s period, likely originating from a 10–15 km long reservoir, lasted for about 20 min (Cesca et al., 2020). Landslides in fjord regions also generate long-period signals, such as the one associated with a tsunami over 100 m high in a Greenland fjord, which produced a monochromatic signal observed globally for 9 days (Carrillo-Ponce et al., 2024; Svennevig et al., 2024). Another notable example is the 26 s signals observed worldwide, originating from around the Gulf of Guinea, off the west coast of Africa and Vanuatu islands, although their physical mechanisms remain uncertain (Bruland & Hadziioannou, 2023; Shapiro et al., 2006). The 26 s period monochromatic signal, located in the Gulf of Guinea, have been identified since the 1960  s (Oliver, 1962). These signals likely originate from volcanic or oceanic processes, with the 26 s microseism in the Vanuatu region possibly linked to the activity of the Ambrym volcano (Kawano et al., 2020). Characterizing these monochromatic signals may provide valuable insights into volcanic activity, ocean waves, and glacier dynamics. However, they often remain undetected by global surface wave-based methods due to their narrow frequency bands, long duration, and unclear onset.

In this study, we apply a coherence-based approach to characterize long-period monochromatic seismic signals. We compute temporal coherence averaged across all station pairs in a regional seismic array in Japan from 2003 to 2022. Through our analysis, we identify several periods during which coherent and long-lasting monochromatic signals occurred. We discuss in detail both known and unknown signals originating from the Gulf of Guinea, Vanuatu, the Fukutoku-Okanoba submarine volcano, and the Canadian Arctic Islands. We then estimate the source location using inter-station arrival times derived from cross-correlations between global seismic stations. Additionally, using the matched filtering method, we search the repetition of events. Our analysis lays the groundwork for the systematic detection and characterization of volcanic and environmental signals, which are increasing in response to ongoing climate change (Carrillo-Ponce et al., 2024; Svennevig et al., 2024).

Data and Methodology

With the aim of detecting and characterizing long-lasting signals at long periods ($${ >} 10$$ s), generated on a global scale, we developed a series of data analysis tools, building on previous works (Droznin et al., 2015; Gibbons & Ringdal, 2006; Poli, 2024; Ross et al., 2019; Seydoux et al., 2016; Soubestre et al., 2018). These tools were applied to signals recorded by 72 broadband seismic stations of the F-net network in Japan (Figure 2c, Figure S1 in Supporting Information S1). We collected continuous data spanning from 1 January 2004, to 31 December 2022, and the analysis was conducted on a daily basis. We processed 26 hr of data for each day, starting from midnight ($${-}1$$ hr) of a given day to the beginning of the following day ($${+}1$$ hr). Initially, the instrument response at each station was corrected to obtain velocity data. The data were then resampled at 1 Hz for efficient processing. For simplicity, we restricted our analysis to vertical-component seismograms.

This dataset was then used to calculate the network-averaged coherence, following the methodology of previous studies (Seydoux et al., 2016; Soubestre et al., 2018). The frequency-dependent cross-spectral density $$(C)$$ between stations $$(i,j)$$ starting at time $$T$$ is obtained as the ensemble average of the cross-spectra over $$M$$ time windows (Bendat & Piersol, 2011): 1 $${C}_{ij}^{T}\left(f\right)=\frac{1}{M}\frac{{\sum }_{t=1}^{M}{S}_{i}\left(t,f\right)\cdot {S}_{j}^{\ast }\left(t,f\right)}{{\sum }_{t=1}^{M}\sqrt{{S}_{i}^{2}\left(t,f\right)\cdot {S}_{j}^{2}\left(t,f\right)}}$$

The size of the time window $$t$$ is 600 s, and the signals $$(S)$$ are tapered using a Hanning window and then whitened in the frequency domain. With $$M=12$$, we averaged over 2 hr, using consecutive windows with 50% overlap. We stored the averaged cross-spectral density, which was later used for beamforming (Rost & Thomas, 2002) and matched filtering (Droznin et al., 2015). For visualization purposes we defined the network-averaged coherence as: 2 $$Co{h}_{\mathit{net}}(T,f)=\frac{2}{N(N-1)}\sum\limits _{i=1}^{N}\sum\limits _{j=i+1}^{N}\vert {C}_{ij}^{T}(f)\vert $$

For signals of interest, we additionally estimated the direction $$({\Theta })$$ and slowness $$(s)$$ of incoming waves using plane wave beamforming (Rost & Thomas, 2002): 3 $$B(f,T)={b}^{T}(f,{\Theta },s){C}_{ij}^{T}b(f,{\Theta },s),$$ where $$b$$ is the steering vector describing the propagation of a plane wave across the array.

To identify repetitions of detected signals, we performed a matched filtering analysis (Droznin et al., 2015), by calculating the cosine distance in between the complex valued coherence matrices estimated at different times, $${T}_{\mathrm{det}}={T}_{1},{T}_{2}$$: 4 $${C}_{\mathrm{det}}\left({T}_{\mathrm{det}}\right)=\frac{1}{m}\sum\limits _{f1}^{f2}\left\vert \frac{{\sum }_{i=1}^{N}{C}_{{T}_{1}}^{\ast }\cdot {C}_{{T}_{2}}}{\sqrt{{\sum }_{i=1}^{N}\vert {C}_{{T}_{1}}{\vert }^{2}}\sqrt{{\sum }_{i=1}^{N}\vert {C}_{{T}_{2}}{\vert }^{2}}}\right\vert ,$$ where $${f}_{1}$$ and $${f}_{2}$$ represent the limits of a frequency band with $$m$$ points, and the index $$i$$ indicates the station couples. The value of $${C}_{\mathrm{det}}$$ takes values between 0 and 1, indicating completely dissimilar and identical signals, respectively.

We lastly apply a method to locate the detected signals (Droznin et al., 2015), extending our analysis to global seismological networks, and grid searching the position maximizing the coherence as: 5 $$G(x,y)=\sum\limits _{i=1}^{N}\sum\limits _{j=i+1}^{N}{C}_{i,j}\left[{\tau }^{i}(x,y)-{\tau }^{j}(x,y)\right],$$ where $$(x,y)$$ is the hypothetical source location, $$i$$ and $$j$$ are station indexes, $$N$$ is the total number of station pairs, $$\tau $$ is the travel time between stations and the source. We determine the travel time by the maximum value of the envelope of the cross-correlation function. We collected vertical-component seismograms at long-period (LH) and broad-band (BH) stations (Figure S1 in Supporting Information S1) from the Global Seismographic Network (Scripps Institution of Oceanography, 1986) and GEOSCOPE global network (Institut de physique du globe de Paris (IPGP) and École et Observatoire des Sciences de la Terre de Strasbourg (EOST), 1982) for the time period of interest. Equation 5 with global stations are used solely to locate signals detected by Equation 2. To estimate the source location, we selected correlations with a signal-to-noise ratio exceeding five times the average value.

Result

The application of Equations 1 and 2 to data recorded by the F-net array from 1 January 2003, to 31 December 2022, resulted in the estimation of over 320,000 network-average coherence vectors within the frequency range of 0.0025–0.5 Hz. These coherence estimates captured a wide range of signals, from earthquakes—characterized by time-localized, broadband increases in coherence—to long-duration microseismic signals, typically confined to narrower frequency bands, lasting up to several months (Figure 1).

[IMAGE OMITTED. SEE PDF]

26-Second Microseisms

To demonstrate the efficacy of our method in tracking the evolution of long-lasting signals, we begin by focusing on some known seismic signals. Specifically, we studied the 26 s microseismic peaks (Bruland & Hadziioannou, 2023; Kawano et al., 2020; Oliver, 1962; Shapiro et al., 2006; Zeng & Ni, 2014) originating from either the Gulf of Guinea or near the Vanuatu Islands (Bruland & Hadziioannou, 2023; Kawano et al., 2020; Oliver, 1962; Shapiro et al., 2006; Zeng & Ni, 2014). To this end, we first visually identified a period during which these signals are strongly observed in our coherence plots (Figures 2a and 2b). The first signal (Figure 2a) peaks at approximately 0.04 Hz and lasts for nearly 24 hr, preceded by a dispersive signal (Bruland & Hadziioannou, 2023). The application of beamforming (Equation 3) to the F-net data reveals a source azimuth consistent with the expected source for the 26 s peak around the Gulf of Guinea (Figure 2a).

[IMAGE OMITTED. SEE PDF]

We performed the same analysis for a 10 day monochromatic signal, which also peaks at approximately 0.04 Hz (Figure 2b). The resolved azimuth from beamforming suggests a source region near the Vanuatu Islands, aligning with previous studies (Kawano et al., 2020; Shapiro et al., 2006; Zeng & Ni, 2014). The spectral peaks of the signals from the Vanuatu Islands and the Gulf of Guinea differ slightly: the Vanuatu signal predominates between 24 and 25 s, while the Gulf of Guinea signal peaks around 26 s. Focusing on the source azimuth estimated from beamforming, we searched a position that maximized coherence (Equation 5) using global seismic networks over a grids: latitudes -50°N–70°N and longitudes -120°E-50°E for the Gulf of Guinea; and latitudes -50°N–70°N and longitudes 100°E–220°E for the Vanuatu Arc, with a 0.5° grid resolution (Figures 3a and 3b). The location of each of the estimated sources is consistent with the locations obtained in previous studies. To verify the wave propagation from the estimated location, we plotted cross-correlations against the differential distance from the source to station pairs. The resulting wave packet propagated at approximately 3 km/s, matching the expected group velocity from the beamforming analysis. These two examples (Figures 2a and 2b) clearly demonstrate the power of array coherence in detecting weak, monochromatic, and persistent signals. The beamformer output in the slowness domain shows a ring-like feature, indicating energy arriving from all directions at approximately 3 km/s (Figures 2a and 2b). This ring-like feature likely represents the arrival of Rayleigh waves generated by primary microseisms, whose sources are typically distributed globally (Gualtieri et al., 2019).

[IMAGE OMITTED. SEE PDF]

Next, we employed the wave propagation properties encoded in the array coherence to study the recurrence of these signals. Specifically, we applied Equation 4 to calculate the similarity between the reference signals (Figures 2a and 2b) and the coherence estimated for every time window from 2003 to 2022. The resulting similarity traces (Figures 4a and 4b) highlight several peaks with correlations above the noise level. Using these traces (Figures 4a and 4b), we extract a catalog of events, retaining only the time windows with correlations greater than twice the median for the entire period. To improve the signal-to-noise ratio, we defined a new reference trace by stacking the newly detected events. We then recalculated the similarity between a new reference trace and the each coherence estimate. The processing with this threshold provided 765, and 2,644 newly detected events in the Gulf of Guinea and Vanuatu region, respectively. 26 s microseisms in the Gulf of Guinea were detected regularly throughout the observation period, with higher occurrences during the Southern Hemisphere winter. This frequency distribution is consistent with the observation that amplitudes of 26 s microseisms increase during the Southern Hemisphere winter (Holcomb, 1998; Shapiro et al., 2006).

[IMAGE OMITTED. SEE PDF]

Unlike events in the Gulf of Guinea, which are detected regularly, the number of 26 s microseisms detected in the Vanuatu Arc showed increased bursts in 2004, 2011, and 2016 compared to other years. This rapid increase likely arises from the persistence of 26 s microseisms in the region over several months. The cause of these prolonged signals remains unclear, but it may be related to the origin of the 26 s signals from active volcanoes within the Vanuatu Arc. A previous study analyzed data from a broadband seismometer installed at Ambrym volcano, revealing that these signals originate from the crater's direction (Kawano et al., 2020). Understanding the mechanism of 26 s microseisms in Vanuatu requires analyzing their spatio-temporal evolution, clustering waveforms, and comparing these patterns with volcanic activities in the arc.

Very Long-Period Signals From the Submarine Volcano

We now focus on a period of enhanced coherence on 14 August 2021, within the frequency band of 0.01–0.05 Hz (Figure 2c). Beamforming projections indicated that the signal direction aligns with the region of the Fukutoku-Okanoba volcano, a submarine volcano in Japan's Bonin Islands. This volcano experienced significant eruptions from August 13 to 15, 2021, producing a high eruption plume and ejecting substantial pumice (Maeno et al., 2022). Notably, nearby seismic and geodetic measurements were limited. This eruption series is estimated to have a Volcanic Explosivity Index of 4. On August 14, intermittent phreatomagmatic explosions occurred, resulting in the formation of a new emergent island. While the closest broadband seismic station did not display clear spectral harmonics related to the eruptions, our array coherence revealed a series of peaks at 0.023 Hz, 0.032 Hz, 0.042 Hz, 0.052 Hz, 0.057 Hz, and 0.066 Hz (Figure S2 in Supporting Information S1). The application of signal detection for this event (Equation 4) did not reveal any repetitions (Figure 4c).

One plausible mechanism for monochromatic tremor with multiple spectral peaks during volcanic activity is the oscillation of fluid-filled cracks (Chouet, 1986). We briefly estimated the crack length and thickness using an analytical model of crack oscillation (Maeda & Kumagai, 2013), assuming a rectangular 2D crack. We estimated a crack geometry with a fundamental mode at 0.023 Hz, assuming a sound speed of 1 km/s in the fluid and a P-wave velocity of 5 km/s in the rock, while varying the fluid-to-rock density ratio. The length and thickness estimates across different frequencies reveal a trade-off between these parameters. By varying the fluid-to-rock density ratio from 0.001 to 0.1, we estimate the crack length to be a few kilometers (Figure S2 in Supporting Information S1). However, the seismic wave velocity and attenuation structure of the Fukutoku-Okanoba volcano have not been revealed, and it remains unclear whether cracks of several kilometers in size actually exist beneath the volcano. Additionally, estimating the size of the magma reservoir below the volcano is challenging, as observing crustal deformation in submarine volcanoes is difficult.

Unidentified Signals From the Northern Canadian Arctic Islands

We next study a previously unidentified signal which occurred on the 24 of January 2014, with a duration of approximately 6 h, and characterized by a narrow band coherence around 0.025 Hz (Figure 2d). This signal is not clearly associated with any significant earthquake. The application of beamforming (Equation 4) indicates a source from north-east of Japan (Figure 2d). To better elucidate the source of this event, we apply the location methods (Equation 2) to global seismological data over a grids of latitudes 40–90°N and longitudes -180-20°E (Figure 3d). The source location is estimated to be around the Canadian Arctic Islands.

We assessed the recurrence of this signal by calculating the similarity over a period of 42–44 s. The application of source detection for this event revealed 97 new events (Figure 4d). The findings suggest that glacier-related signals are occurring not only in Antarctica and Greenland, where glacial seismic activities have been detected by the global surface wave analyses (Ekström, 2006; Poli, 2024), but also in the Canadian Arctic Islands.

The identified monochromatic signals originating from the northern Canadian Arctic Islands are not listed in the catalogs created by global surface wave analyses (Poli, 2024). The weak seismic energy from this region significantly hinders detection through traditional global surface wave analysis, which typically focuses on impulsive signals (Ekström, 2006; Poli, 2024; Shearer, 1994). Even at stations near the estimated source location, the signal's waveform remains unclear in the time domain.

However, our method enhances the detection capabilities by revealing a clear spectral peak around 0.025 Hz lasting for 6 hours in the spectrograms of those stations (Figure S3 in Supporting Information S1), consistent with the coherence analysis conducted in Japan's array. This advancement allows for a more accurate identification of the monochromatic signals, even in challenging conditions. Despite the ambiguous waveforms recorded at broadband stations near the Canadian Arctic Islands, our approach enables us to infer important insights into the physical processes behind the observed monochromatic signal.

One possible mechanism for this monochromatic seismic event is a seismic seiche–a standing wave in a channel of ocean water. A seismic seiche in polar regions may be triggered by a landslide entering seawater around complex terrain (Paris et al., 2019; Svennevig et al., 2024). We briefly estimated the shelf length and water depth based on the theory of seiche. The dominant period of a seiche is formulated as follows (Svennevig et al., 2024): 6 $${T}_{p}=\frac{2L}{n\sqrt{gh}}$$ where $$n$$ is the mode number of the wave, $${T}_{p}$$ is the wave period, $$L$$ is the length of the channel, and $$h$$ is the water depth. Considering only the fundamental mode $$(n=1)$$, we estimated the water depth at the source to range from approximately 500–2,500 m, and the shelf length to range from 1.8 to 4 km (Figure S4 in Supporting Information S1). If a landslide occurred within a fjord, these estimates align with typical fjord dimensions. However, the precise origin of this long-period monochromatic signal remains uncertain, as nearby stations recorded no clear waveform in this study. Further research is needed to estimate the extent of the landslide and pinpoint the fjord where oscillations occurred. To confirm these results, additional broadband stations near the Arctic Islands and satellite data analysis will be essential.

Discussion and Conclusions

The focus of our research was to develop an effective approach for characterizing long-duration monochromatic seismic signals on a global scale. To achieve this, we applied network-averaged coherence analyses to continuous seismic data obtained from a regional array, enabling us to capture a diverse range of signals. Our analyses identified both known and previously unrecognized signals originating from the Gulf of Guinea, the Vanuatu region, the Fukutoku-Okanoba submarine volcano, and the Canadian Arctic Islands (Table S1 in Supporting Information S1). These weak and long-duration signals are challenging to detect using conventional surface wave analysis algorithms, demonstrating the effectiveness of our approach.

In general, we found that surface waves from earthquakes in shallow areas exhibited high coherence within a narrow frequency band. Notably, prior to the signal from the Canadian Arctic region, we observed high coherence at a frequency of 0.026 Hz between 14:00 and 17:00 on 24 January 2014 (Figure 2d). This signal corresponded to a M4.7 event that occurred at a depth of 9 km north of Ascension Island. To systematically detect unidentified monochromatic signals based on our method, it is crucial to distinguish them from surface waves generated by shallow earthquakes.

We located identified signals through grid searching for maximum coherence obtained from global seismological networks. At the Fukutoku-Okanoba volcano, our estimated location aligned with the known volcano position. Additionally, when correlation functions were arranged by the distance of each station pair to the estimated source, we observed the propagation of wave packets at the group velocity of surface waves. This suggests that our method for estimating source locations is effective. However, we noted a significant misfit—over several hundred seconds—between observed and calculated arrival times from the hypothetical source location. This uncertainty may arise from inadequate stacking of correlations of monochromatic signals and deviations from great-circle paths. Our approach assumes surface waves propagate along minor-arc great circles, but complexities due to lateral refraction, scattering, and multi-pathing, resulting from Earth's heterogeneous structure, can introduce systematic errors in source location estimates (Laske & Masters, 1996). Furthermore, we implicitly assumed that the waves radiate isotopically from a point source when determining the source location. If the phase at the source vary with takeoff angle, it will increase the uncertainty in location estimation by our approach.

To detect repetitions of the identified signals, we computed correlation coefficients between reference and short-time windowed cross-spectra. By fixing a narrow period band and utilizing only manually identified signals in the coherence vectors as reference cross-spectra, we established a framework for systematically searching and classifying monochromatic events. This systematic approach has the potential to uncover more long-period signals that remain undiscovered.

Furthermore, our study detected more events in the Vanuatu region compared to the Gulf of Guinea, likely due to the geometric spreading effects in the latter. We anticipate that a seismic network positioned closer to the Gulf of Guinea would facilitate the detection of additional 26 s microseisms, revealing more about their temporal evolution, including seasonal variability, which was already hinted at by the limited detections in this study. Examining such detections with arrays outside Japan and exploring the source mechanisms of these 26 s signals lies beyond the scope of this study.

In conclusion, we successfully characterized long-period monochromatic seismic signals through coherence analysis of continuous seismic data recorded by a broadband seismic network in Japan. These signals exhibit sharp spectral peaks and long durations, which conventional global long-period surface wave analyses do not typically identify. The coherence across all stations in our regional array highlighted long-duration signals from the Gulf of Guinea, the Vanuatu Arc, the Fukutoku-Okanoba submarine volcano, and the Canadian Arctic Islands, as well as microseisms at a period of 14 s. By employing cross-correlation function analysis with a global seismic network, we estimated the locations of these monochromatic signals' origins. To identify repetitions of detected events, we calculated the cosine distance between coherence matrices at different times. Notably, this template matching of coherence matrices suggests new glacial seismic activities in the northern Canadian Arctic region, seasonal occurrences of 26 s signals in the Gulf of Guinea, and prolonged signals from the Vanuatu Arc over several months. Our findings lay the groundwork for the systematic detection and characterization of volcanic and environmental signals, which are increasingly responsive to ongoing climate change (Svennevig et al., 2024). Developing an automated algorithm to detect and locate monochromatic seismic signals will facilitate the discovery of previously unidentified signals, thereby enhancing our understanding of physical processes occurring on the Earth's surface and within its interior.

Acknowledgments

We appreciate all the people who working on building and maintaining seismic networks. We thank Editor Germán Prieto and two anonymous reviewers for their constructive remarks that improved our manuscript. This work is supported by JSPS KAKENHI Grand Number JP22K14110.

Data Availability Statement

We downloaded seismic data through EarthScope Consortium Web Services (), including the following seismic networks: II (Scripps Institution of Oceanography, 1986); IU (Albuquerque Seismological Laboratory/USGS, 2014); G (Institut de physique du globe de Paris (IPGP) & École et Observatoire des Sciences de la Terre de Strasbourg (EOST), 1982). We used F-net data (NIED F-net, 2019). Catalogs of events which we have detected in the Gulf of Guinea and Vanuatu, and the Canadian arctic region are available at the Zenodo webpage (Takano, 2024).

References

Albuquerque Seismological Laboratory/USGS. (2014). Global seismograph network (gsn ‐ Iris/usgs) [dataset]. International Federation of Digital Seismograph Networks. Retrieved from https://www.fdsn.org/networks/detail/IU/doi:10.7914/SN/IU

Beaucé, E., Frank, W. B., & Romanenko, A. (2018). Fast matched filter (fmf): An efficient seismic matched‐filter search for both cpu and gpu architectures. Seismological Research Letters, 89(1), 165–172. https://doi.org/10.1785/0220170181

Beaucé, E., Frank, W. B., Seydoux, L., Poli, P., Groebner, N., van der Hilst, R. D., & Campillo, M. (2024). Bpmf: A backprojection and matched‐filtering workflow for automated earthquake detection and location. Seismological Research Letters, 95(2A), 1030–1042. https://doi.org/10.1785/0220230230

Bendat, J. S., & Piersol, A. G. (2011). Random data: Analysis and measurement procedures. John Wiley and Sons.

Bergen, K. J., Johnson, P. A., de Hoop, M. V., & Beroza, G. C. (2019). Machine learning for data‐driven discovery in solid earth geoscience. Science, 363(6433), eaau0323. https://doi.org/10.1126/science.aau0323

Bruland, C., & Hadziioannou, C. (2023). Gliding tremors associated with the 26 second microseism in the gulf of Guinea. Communications Earth and Environment, 4(1), 176. https://doi.org/10.1038/s43247‐023‐00837‐y

Carrillo‐Ponce, A., Heimann, S., Petersen, G. M., Walter, T. R., Cesca, S., & Dahm, T. (2024). The 16 september 2023 Greenland megatsunami: Analysis and modeling of the source and a week‐long, monochromatic seismic signal. The Seismic Record, 4(3), 172–183. https://doi.org/10.1785/0320240013

Cesca, S., Letort, J., Razafindrakoto, H. N., Heimann, S., Rivalta, E., Isken, M. P., et al. (2020). Drainage of a deep magma reservoir near Mayotte inferred from seismicity and deformation. Nature Geoscience, 13(1), 87–93. https://doi.org/10.1038/s41561‐019‐0505‐5

Chen, X., Shearer, P., Walter, F., & Fricker, H. (2011). Seventeen antarctic seismic events detected by global surface waves and a possible link to calving events from satellite images. Journal of Geophysical Research, 116(B6), B06311. https://doi.org/10.1029/2011jb008262

Chouet, B. (1986). Dynamics of a fluid‐driven crack in three dimensions by the finite difference method. Journal of Geophysical Research, 91(B14), 13967–13992. https://doi.org/10.1029/jb091ib14p13967

Chouet, B. (2003). Volcano seismology. Pure and Applied Geophysics, 160, 739–788. https://doi.org/10.1007/978‐3‐0348‐8010‐7_13

Droznin, D., Shapiro, N., Droznina, S. Y., Senyukov, S., Chebrov, V., & Gordeev, E. (2015). Detecting and locating volcanic tremors on the klyuchevskoy group of volcanoes (kamchatka) based on correlations of continuous seismic records. Geophysical Journal International, 203(2), 1001–1010. https://doi.org/10.1093/gji/ggv342

Ekström, G. (2006). Global detection and location of seismic sources by using surface waves. Bulletin of the Seismological Society of America, 96(4A), 1201–1212. https://doi.org/10.1785/0120050175

Essing, D., & Poli, P. (2022). Spatiotemporal evolution of the seismicity in the alto tiberina fault system revealed by a high‐resolution template matching catalog. Journal of Geophysical Research: Solid Earth, 127(10), e2022JB024845. https://doi.org/10.1029/2022jb024845

Fan, W., de Groot‐Hedlin, C. D., Hedlin, M. A., & Ma, Z. (2018). Using surface waves recorded by a large mesh of three‐element arrays to detect and locate disparate seismic sources. Geophysical Journal International, 215(2), 942–958. https://doi.org/10.1093/gji/ggy316

Fan, W., McGuire, J. J., & Shearer, P. M. (2020). Abundant spontaneous and dynamically triggered submarine landslides in the gulf of Mexico. Geophysical Research Letters, 47(12), e2020GL087213. https://doi.org/10.1029/2020gl087213

Gibbons, S. J., & Ringdal, F. (2006). The detection of low magnitude seismic events using array‐based waveform correlation. Geophysical Journal International, 165(1), 149–166. https://doi.org/10.1111/j.1365‐246x.2006.02865.x

Gualtieri, L., Stutzmann, E., Juretzek, C., Hadziioannou, C., & Ardhuin, F. (2019). Global scale analysis and modelling of primary microseisms. Geophysical Journal International, 218(1), 560–572. https://doi.org/10.1093/gji/ggz161

Holcomb, L. G. (1998). Spectral structure in the earth’s microseismic background between 20 and 40 seconds. Bulletin of the Seismological Society of America, 88(3), 744–757. https://doi.org/10.1785/bssa0880030744

Institut de physique du globe de Paris (IPGP), and École et Observatoire des Sciences de la Terre de Strasbourg (EOST). (1982). Geoscope, French global network of broad band seismic stations [dataset]. Institut de physique du globe de Paris (IPGP). Retrieved from https://doi.org/10.18715/GEOSCOPE.G

Kawano, Y., Isse, T., Takeo, A., Kawakatsu, H., Suetsugu, D., Shiobara, H., et al. (2020). Persistent long‐period signals recorded by an obs array in the western‐central pacific: Activity of ambrym volcano in Vanuatu. Geophysical Research Letters, 47(19), e2020GL089108. https://doi.org/10.1029/2020gl089108

Kong, Q., Trugman, D. T., Ross, Z. E., Bianco, M. J., Meade, B. J., & Gerstoft, P. (2019). Machine learning in seismology: Turning data into insights. Seismological Research Letters, 90(1), 3–14. https://doi.org/10.1785/0220180259

Kumagai, H., & Chouet, B. A. (2000). Acoustic properties of a crack containing magmatic or hydrothermal fluids. Journal of Geophysical Research, 105(B11), 25493–25512. https://doi.org/10.1029/2000jb900273

Laske, G., & Masters, G. (1996). Constraints on global phase velocity maps from long‐period polarization data. Journal of Geophysical Research, 101(B7), 16059–16075. https://doi.org/10.1029/96jb00526

Lin, G., & Shearer, P. (2006). The comploc earthquake location package. Seismological Research Letters, 77(4), 440–444. https://doi.org/10.1785/gssrl.77.4.440

Maeda, Y., & Kumagai, H. (2013). An analytical formula for the longitudinal resonance frequencies of a fluid‐filled crack. Geophysical Research Letters, 40(19), 5108–5112. https://doi.org/10.1002/grl.51002

Maeno, F., Kaneko, T., Ichihara, M., Suzuki, Y. J., Yasuda, A., Nishida, K., & Ohminato, T. (2022). Seawater‐magma interactions sustained the high column during the 2021 phreatomagmatic eruption of fukutoku‐oka‐no‐ba. Communications Earth and Environment, 3(1), 260. https://doi.org/10.1038/s43247‐022‐00594‐4

Majstorović, J., Giffard‐Roisin, S., & Poli, P. (2021). Designing convolutional neural network pipeline for near‐fault earthquake catalog extension using single‐station waveforms. Journal of Geophysical Research: Solid Earth, 126(7), e2020JB021566. https://doi.org/10.1029/2020jb021566

McGuire, J. J. (2008). Seismic cycles and earthquake predictability on east pacific rise transform faults. Bulletin of the Seismological Society of America, 98(3), 1067–1084. https://doi.org/10.1785/0120070154

Meng, X., & Peng, Z. (2014). Seismicity rate changes in the salton sea geothermal field and the san jacinto fault zone after the 2010 m w 7.2 el mayor‐cucapah earthquake. Geophysical Journal International, 197(3), 1750–1762. https://doi.org/10.1093/gji/ggu085

Mousavi, S. M., Ellsworth, W. L., Zhu, W., Chuang, L. Y., & Beroza, G. C. (2020). Earthquake transformer—An attentive deep‐learning model for simultaneous earthquake detection and phase picking. Nature Communications, 11(1), 3952. https://doi.org/10.1038/s41467‐020‐17591‐w

NIED F‐net. (2019). F‐net broadband seismograph network. National Research Institute for Earth Science and Disaster Resilience. [dataset]. https://doi.org/10.17598/NIED.0005

Oliver, J. (1962). A worldwide storm of microseisms with periods of about 27 seconds. Bulletin of the Seismological Society of America, 52(3), 507–517. https://doi.org/10.1785/bssa0520030507

Paris, A., Okal, E. A., Guérin, C., Heinrich, P., Schindelé, F., & Hébert, H. (2019). Numerical modeling of the june 17, 2017 landslide and tsunami events in karrat fjord, west Greenland. Pure and Applied Geophysics, 176(7), 3035–3057. https://doi.org/10.1007/s00024‐019‐02123‐5

Poli, P. (2024). Continuation of events detection with global long‐period seismic data: An analysis from 2010 to 2022. Seismological Research Letters, 95(3), 1486–1494. https://doi.org/10.1785/0220230148

Ross, Z. E., Trugman, D. T., Hauksson, E., & Shearer, P. M. (2019). Searching for hidden earthquakes in southern California. Science, 364(6442), 767–771. https://doi.org/10.1126/science.aaw6888

Rost, S., & Thomas, C. (2002). Array seismology: Methods and applications. Reviews of Geophysics, 40(3), 1–2. https://doi.org/10.1029/2000rg000100

Scripps Institution of Oceanography. (1986). Global seismograph network ‐ IRIS/IDA. Retrieved from https://www.fdsn.org/networks/detail/II/10.7914/SN/II

Seydoux, L., Shapiro, N. M., de Rosny, J., Brenguier, F., & Landès, M. (2016). Detecting seismic activity with a covariance matrix analysis of data recorded on seismic arrays. Geophysical Journal International, 204(3), 1430–1442. https://doi.org/10.1093/gji/ggv531

Shapiro, N. M., Ritzwoller, M., & Bensen, G. (2006). Source location of the 26 sec microseism from cross‐correlations of ambient seismic noise. Geophysical Research Letters, 33(18). https://doi.org/10.1029/2006gl027010

Shearer, P. M. (1994). Global seismic event detection using a matched filter on long‐period seismograms. Journal of Geophysical Research, 99(B7), 13713–13725. https://doi.org/10.1029/94jb00498

Soubestre, J., Shapiro, N. M., Seydoux, L., de Rosny, J., Droznin, D. V., Droznina, S. Y., et al. (2018). Network‐based detection and classification of seismovolcanic tremors: Example from the klyuchevskoy volcanic group in kamchatka. Journal of Geophysical Research: Solid Earth, 123(1), 564–582. https://doi.org/10.1002/2017jb014726

Svennevig, K., Hicks, S. P., Forbriger, T., Lecocq, T., Widmer‐Schnidrig, R., Mangeney, A., et al. (2024). A rockslide‐generated tsunami in a Greenland fjord rang earth for 9 days. Science, 385(6714), 1196–1205. https://doi.org/10.1126/science.adm9247

Takano, T. (2024). tomoyatakano/catalog‐of‐monochromatic‐signal: Ver1.0.0 (ver1.0.0). Zenodo. [dataset]. https://doi.org/10.5281/zenodo.14402696

Thurber, C. H. (1983). Earthquake locations and three‐dimensional crustal structure in the coyote lake area, central California. Journal of Geophysical Research, 88(B10), 8226–8236. https://doi.org/10.1029/jb088ib10p08226

Thurber, C. H. (1985). Nonlinear earthquake location: Theory and examples. Bulletin of the Seismological Society of America, 75(3), 779–790. https://doi.org/10.1785/bssa0750030779

Tsai, V. C., & Ekström, G. (2007). Analysis of glacial earthquakes. Journal of Geophysical Research, 112(F3). https://doi.org/10.1029/2006jf000596

Waldhauser, F., & Ellsworth, W. L. (2000). A double‐difference earthquake location algorithm: Method and application to the northern hayward fault, California. Bulletin of the Seismological Society of America, 90(6), 1353–1368. https://doi.org/10.1785/0120000006

Wilding, J. D., Zhu, W., Ross, Z. E., & Jackson, J. M. (2023). The magmatic web beneath hawai ‘i. Science, 379(6631), 462–468. https://doi.org/10.1126/science.ade5755

Yamamoto, M., Kawakatsu, H., Kaneshima, S., Mori, T., Tsutsui, T., Sudo, Y., & Morita, Y. (1999). Detection of a crack‐like conduit beneath the active crater at aso volcano Japan. Geophysical Research Letters, 26(24), 3677–3680. https://doi.org/10.1029/1999gl005395

Yukutake, Y., Kim, A., & Ohminato, T. (2023). Reappraisal of volcanic seismicity at the kirishima volcano using machine learning. Earth Planets and Space, 75(1), 183. https://doi.org/10.1186/s40623‐023‐01939‐9

Zeng, X., & Ni, S. (2014). Evidence for an independent 26‐s microseismic source near the Vanuatu islands. Pure and Applied Geophysics, 171(9), 2155–2163. https://doi.org/10.1007/s00024‐014‐0811‐1