Seismo-volcanic Signals Research Articles

Recurrent Scattering Network Detects Metastable Behavior in Polyphonic Seismo-Volcanic Signals for Volcano Eruption Forecasting

Recurrent Scattering Network Detects Metastable Behavior in Polyphonic Seismo-Volcanic Signals for Volcano Eruption Forecasting

A machine-learning approach for automatic classification of volcanic seismicity at La Soufrière Volcano, Guadeloupe

The classification of seismo-volcanic signals is performed manually at La Soufrière Volcano, which is time consuming and can be biased by subjectivity of the operator. We propose here a machine-learning-based model for classification of these signals, to handle large datasets and provide objective and reproducible results. To describe the properties of the signals, we used 104 statistical, entropy, and shape descriptor features computed from the time waveform, the spectrum, and the cepstrum. First, we trained a random forest classifier with a dataset provided by the Observatoire Volcanologique et Sismologique de Guadeloupe that consisted of 845 labeled events that were recorded from 2013 to 2018: 542 volcano-tectonic (VT); 217 Nested; and 86 long period (LP). We obtained an overalll accuracy of 72%. We determined that the VT class includes a variety of signals that cover the VT, Nested and LP classes. After visual inspection of the waveforms and spectral characteristics of the data set, we introduced two new classes: Hybrid and Tornillo. A new random forest classifier was trained with this new information, and we obtained a much better overall accuracy of 82%. The model is very good for recognition of all event classes, except Hybrid events (67% accuracy, 70% precision). Hybrid events are often considered to be a mix of VT and LP events. This can be explained by the nature of this class and the physical processes that include both fracturing and resonating components with different modal frequencies. By analyzing the feature weights and by training a model with the most important features, we show that a subset of the 14 best features is sufficient to obtain a performance that is close to that of the model with the whole feature set. However, these best features are different from the 13 best features obtained for another volcano in Peru, with only one feature common to both sets of best features. Therefore, the model is not universal and it must be trained for each volcano, or it is too specific to the one station used here.

Detection, Classification, and Location of Seismovolcanic Signals with Multicomponent Seismic Data: Example from the Piton de La Fournaise Volcano (La Réunion, France)

AbstractWe apply three different methods based on the analysis of the multicomponent seismic data to detect seismovolcanic tremors and other seismovolcanic signals, to propose an approach to classify them, and to locate their sources. We use continuous seismograms recorded during 1 year by 21 stations at the Piton de la Fournaise volcano (La Réunion, France). The first method allows the detection of seismovolcanic signals based on stability in time of the intercomponent cross‐correlation function. Two other methods based on the simultaneous analysis of the whole network can be used to detect seismovolcanic signals and to locate their sources. In the first network‐based method, the seismic wavefield is analyzed by calculating the width of the network covariance matrix eigenvalue distribution.The second network‐based method consists in performing the 3‐D backprojection of the interstation cross correlations in order to calculate the network response function. Simultaneous analysis of the parameters measured by the three different methods can be used to classify different types of seismovolcanic tremors. Our results demonstrate that all three methods efficiently detect seismovolcanic tremors accompanying the 2010 eruptions and the preceding pre‐eruptive seismic swarms. Furthermore, Methods 2 and 3 based on simultaneous analysis of the whole network detect a large number of volcanic earthquakes. Our location results show that each seismovolcanic tremor is located in a distinct regionof the volcano, close to the eruptive site at a shallow depth, and the preceding seismic crisis is located deeper at about the sea level under the summit crater.

Genesis and mechanisms controlling tornillo seismo-volcanic events in volcanic areas

Volcanic activity is often preceded or accompanied by different types of seismo-volcanic signals. Among these signals, the so-called tornillo (Spanish for “screw”) events are considered to belong to a unique class of volcano-seismicity characterised by a long-duration coda, amplitude modulation and high-quality factor. These data constitute important evidence for the gas fraction inside magmatic fluids. However, the mechanism behind this unique signal remains not fully understood. Here we report new laboratory evidence showing that two different processes have either scale-invariant or scale-dependent effects in generating tornillo-like events. These processes are respectively the gas pressure gradient, which triggers the event and regulates the slow decaying coda, and the fluid resonance into small scale structures which, in turn, control the frequency content of the signal. Considering that the gas pressure gradient is proportional to the fluid flow, these new findings, as applied to volcanoes, provide new information to better quantify both gas rate and volume, and the dimension of the resonator.

Localizando señales sismo- volcánicas del volcán Turrialba (Costa Rica) usando Python y Computación Avanzada: Un caso de colaboración multidisciplinar para el desarrollo cientí co

Las señales sismo-volcánicas de baja frecuencia son generadas por el movimiento interno de magma, gases, agua, entre otros. Suelen preceder o acompañar la actividad eruptiva. Por lo tanto, estudiarlos resulta fundamental para el monitoreo de la actividad volcánica y una apropiada estimación de la amenaza. Sin embargo, la complejidad del mecanismo de la fuente hace que los procedimientos clásicos de localización de sismos generados por procesos tectónicos, que son los más comunes, no puedan ser utilizados para los volcánicos. A esto se suma la heterogeneidad de los edificios volcánicos que modifican en gran medida las formas de onda de las señales sismo-volcánicas, lo que también dificulta su procesamiento. Es por ello que se requiere aplicar otro tipo de métodos de tratamiento de señal los cuales, por su complejidad computacional, requieren de plataformas de computación avanzada (supercomputadoras), sobretodo cuando se precisa de resultados en tiempo real. No obstante, los observatorios o entes encargados de la auscultación y monitoreo volcánico en la región centroamericana tienen recursos limitados y no pueden mantener un departamento completo de desarrollo de software especializado. La presente ponencia analiza un caso de estudio, en el cual se desarrolló un trabajo colaborativo entre especialistas en sismología volcánica y en computación avanzada, quienes implementaron una plataforma computacional para la localización de señales sismo- volcánicas en el Volcán Turrialba, Costa Rica. Nuestra principal conclusión es que la creación de redes de colaboración multidisciplinaria es una opción que permite maximizar recursos para abordar y superar muchas de las limitaciones que existen para el desarrollo de la investigación en nuestro contexto (falta de recursos humano, tecnología, presupuesto, entre otros).

Advances in seismic monitoring at Deception Island volcano (Antarctica) since the International Polar Year

<p>Deception Island is an active volcano located in the south Shetland Islands, Antarctica. It constitutes a natural laboratory to test geophysical instruments in extreme conditions, since they have to endure not only the Antarctic climate but also the volcanic environment. Deception is one of the most visited places in Antarctica, both by scientists and tourists, which emphasize the importance of volcano monitoring. Seismic monitoring has been going on since 1986 during austral summer surveys. The recorded data include volcano-tectonic earthquakes, long-period events and volcanic tremor, among others. The level of seismicity ranges from quiet periods to seismic crises (e.g. 1992-1993, 1999). Our group has been involved in volcano monitoring at Deception Island since 1994. Based on this experience, in recent years we have made the most of the opportunities of the International Polar Year 2007-2008 to introduce advances in seismic monitoring along four lines: (1) the improvement of the seismic network installed for seismic monitoring during the summer surveys; (2) the development and improvement of seismic arrays for the detection and characterization of seismo-volcanic signals; (3) the design of automated event recognition tools, to simplify the process of data interpretation; and (4) the deployment of permanent seismic stations. These advances help us to obtain more data of better quality, and therefore to improve our interpretation of the seismo-volcanic activity at Deception Island, which is a crucial step in terms of hazards assessment.</p>

The classification of seismo-volcanic signals using Hidden Markov Models as applied to the Stromboli and Etna volcanoes

The aim of this work is to apply the Hidden Markov Model (HMM) method to recognise seismic signals belonging to different active volcanoes. We use data obtained from two field surveys carried out in 1997 and 1999 at Stromboli and Etna, respectively. For Stromboli we used two types of seismic signals for recognition purposes: Strombolian explosions and background seismic noise whilst for Etna we used volcanic tremors and tremor bursts. We initially proceeded to visually identify the signals, and to segment the data to obtain a model for each event class. We then applied these models separately to each volcano dataset, finally combining both datasets as a test of the portability of the system. The method analyses the seismograms and compares the characteristics of the data to a number of pre-defined event classes. If a signal is present, the method detects its occurrence and produces a classification. We observed that, to obtain reliable results, a careful and adequate segmentation process is crucial and that each signal type must be represented by its own specific model. Once we had built this model, the success level of the system was high. The success rates obtained indicated that the method was fully able to detect, isolate, and identify signals from raw seismic data. These results imply that, once an adequate training process has been used, this method is particularly appropriate for work in real time, as well as concurrently with the data acquisition system.

Coherent-subspace array processing based on wavelet covariance: an application to broad-band, seismo-volcanic signals

SUMMARY Long-Period (LP) and Very-Long-Period (VLP) signals are the most characteristic seismic signature of volcano dynamics, and provide important information about the physical processes occurring in magmatic and hydrothermal systems. These events are usually characterized by sharp spectral peaks, which may span several frequency decades, by emergent onsets, and by a lack of clear S-wave arrivals. These two latter features make both signal detection and location a challenging task. In this paper, we propose a processing procedure based on Continuous Wavelet Transform of multichannel, broad-band data to simultaneously solve the signal detection and location problems. Our method consists of two steps. First, we apply a frequency-dependent threshold to the estimates of the array-averaged WCO in order to locate the time-frequency regions spanned by coherent arrivals. For these data, we then use the time-series of the complex wavelet coefficients for deriving the elements of the spatial Cross-Spectral Matrix. From the eigenstructure of this matrix, we eventually estimate the kinematic signals' parameters using the MUltiple SIgnal Characterization (MUSIC) algorithm. The whole procedure greatly facilitates the detection and location of weak, broad-band signals, in turn avoiding the time-frequency resolution trade-off and frequency leakage effects which affect conventional covariance estimates based upon Windowed Fourier Transform. The method is applied to explosion signals recorded at Stromboli volcano by either a short-period, small aperture antenna, or a large-aperture, broad-band network. The LP (0.2 2 s) of the explosion recordings from the broad-band network. Source locations obtained this way are fully compatible with those retrieved from application of more traditional (and computationally expensive) time-domain techniques, such as the Radial Semblance method.

The recent seismo-volcanic activity at Deception Island volcano

Abstract This paper reviews the recent seismic studies carried out at Deception Island, South Shetland Islands, Antarctica, which was monitored by the Argentinean and Spanish Antarctic Programs since 1986. Several types of seismic network have been deployed temporarily during each Antarctic summer. These networks have consisted of a variety of instruments, including radio-telemetered stations, autonomous digital seismic stations, broadband seismometers, and seismic arrays. We have identified two main types of seismic signals generated by the volcano, namely pure seismo-volcanic signals, such as volcanic tremor and long-period (LP) events, and volcano-tectonic (VT) earthquakes. Their temporal distributions are far from homogeneous. Volcanic tremors and LP events usually occur in seismic swarms lasting from a few hours to some days. The number of LP events in these swarms is highly variable, from a background level of less than 30/day to a peak activity of about 100 events/h. The occurrence of VT earthquakes is even more irregular. Most VT earthquakes at Deception Island have been recorded during two intense seismic crises, in 1992 and 1999, respectively. Some of these VT earthquakes were large enough to be felt by researchers working on the island. Analyses of both types of seismic events have allowed us to derive source locations, establish seismic source models, analyze seismic attenuation, calculate the energy and stress drop of the seismic sources, and relate the occurrence of seismicity to the volcanic activity. Pure seismo-volcanic signals are modelled as the consequence of hydrothermal interactions between a shallow aquifer and deeper hot materials, resulting in the resonance of fluid-filled fractures. VT earthquakes constitute the brittle response to changes in the distribution of stress in the volcanic edifice. The two VT seismic series are probably related to uplift episodes due to deep injections of magma that did not reach the surface. This evidence, however, indicates the high potential for future volcanic eruptions at Deception Island.

Seismovolcanic signals at Deception Island volcano, Antarctica: Wave field analysis and source modeling

The seismovolcanic signals associated with the volcanic activity of Deception Island (Antarctica), recorded during three Antarctic summers (1994–1995, 1995–1996 and 1996–1997), are analyzed using a dense small‐aperture (500 m) seismic array. The visual and spectral classification of the seismic events shows the existence of long‐period and hybrid isolated seismic events, and of low‐frequency, quasi‐monochromatic and spasmodic continuous tremors. All spectra have the highest amplitudes in the frequency band between 1 and 4 Hz, while hybrids and spasmodic tremors have also significant amplitudes in the high‐frequency band (4–10 Hz). The array analysis indicates that almost all the well‐correlated low‐frequency signals share similar array parameters (slowness and back azimuth) and have the same source area, close to the array site. The polarization analysis shows that phases at high‐frequency are mostly composed of P waves, and those phases dominated by low frequencies can be interpreted as surface waves. No clear shear waves are evidenced. From the energy evaluation, we have found that the reduced displacement values for surface and body waves are confined in a narrow interval. Volcano‐tectonic seismicity is located close to the array, at a depth shallower than 1 km. The wave‐field properties of the seismovolcanic signals allow us to assume a unique source model, a shallow resonating fluid‐filled crack system at a depth of some hundreds of meters. All of the seismic activity is interpreted as the response of a reasonably stable stationary geothermal process. The differences observed in the back azimuth between low and high frequencies are a near‐field effect. A few episodes of the degassification process in an open conduit were observed and modeled with a simple organ pipe.