Multi-horizon Simultaneous Tracking Based on Dynamic Seismic Waveform Matching

Summary Due to the rapid increase in the amount of seismic volumes, the traditional seismic interpretation mode based on manual structure interpretation and single-horizon automatic tracking has encountered many challenges. The seismic interpretation of large or super-large 3-D seismic surveys is facing serious accuracy and efficiency bottlenecks. Aiming to the goal of improving the accuracy and efficiency of seismic interpretation, we propose a dynamic seismic waveform matching technology based on the sparse dynamic time warping algorithm under the guidance of the relative geological time volume theory, and realize multi-horizon simultaneous tracking based on the technology. Has been verified by a model and a real seismic volume, it can realize simultaneous horizon automatic tracking, full spatial tracking and high-density tracking, and can significantly improve the accuracy and efficiency of structure interpretation.

Adaptive Gaussian Mixture Model and Convolution Autoencoder Clustering for Unsupervised Seismic Waveform Analysis

Seismic facies analysis can effectively estimate reservoir properties and seismic waveform clustering is a useful tool for facies analysis. We developed a deep learning-based clustering approach called the modified deep convolutional embedded clustering with adaptive Gaussian mixture model (AGMM-MDCEC) for seismic waveform clustering. Trainable feature extraction and clustering layers in AGMM-MDCEC are implemented using neural networks. The two independent processes of feature extraction and clustering are fused, such that extracted features are modified simultaneously with the results of clustering. A convolutional autoencoder is used in the algorithm for extracting features from seismic data and reduce data redundancy. At the same time, weights of clustering network are fined-tuned through iteration to obtain state-of-the-art clustering results. We apply our new classification algorithm to a data volume acquired in western China to map architectural elements of a complex fluvial depositional system. Our proposed method obtains superior results over those provided by traditional K-means, Gaussian mixture model, and some machine learning methods, and improves the mapping of the extent of the distributary system.

Microseismicity Associated with Actively Exploited Geothermal Systems: Earthquake Detection and Probabilistic Location at Rotokawa and Statistical Seismic Network Design at Kawerau

<p>Geothermal power has progressively been recognised as an important energy resource due to the depletion of old power sources, and as a more environmentally aware population pushes for an increase in renewable energy sources. Monitoring microseismicity occurring in active geothermal systems is one means of both characterising the system’s fault architecture and characterising fluid/rock interaction in response to production. This study focuses on better understanding seismicity in two active geothermal fields, through the development and implementation of two different algorithms: an automated microearthquake detection algorithm using a matched filter technique (improving earthquake detection), and an optimal seismic network design algorithm (improving earthquake location). Both algorithms have been implemented in codes that are easily adaptable to other data sets. The first of these algorithms has been applied to five months of continuous seismic waveform data spanning a fluid injection operation in the Rotokawa geothermal field. The cross-correlation of 14 high-quality master events with the continuous seismic data yields 2461 newly detected earthquakes spanning the magnitude range M=-0.4 to M=2.6 with a mean magnitude of M=0.47. The earthquakes detected with each master event exhibit high waveform similarity over approximately three orders of magnitude, and appear to follow a Gutenberg-Richter power law with a catalogue completeness down to M~ 0. Hypocentres for these detected events computed using the probabilistic earthquake location algorithm NonLinLoc reveal the dominant locus of seismicity to lie between 1.0–2.5 km depth, a location consistent with that of the Rotokawa Andesite which forms the Rotokawa reservoir. Focal mechanism solutions for the master events are predominantly normal, with half displaying a large strike-slip component, and the stress parameters obtained for this suite of focal mechanisms imply a northeast–southwest oriented maximum horizontal stress: both of these results are consistent with the extensional regime of the TVZ. Seismicity occurring within a 300 m horizontal radius of the injection well’s feed-zones, and extending to 5 km depth, initially exhibits a correlation with injection flow rates with a ~ 2 day lag, and seismicity rates decrease ~ 10 weeks after injection. We surmise that seismicity within the injection region and close to the injection well is likely to be injection-induced, with one portion of the injectate returning to the production region, while the other either migrates southeastward out of the field or remains within the injection region; the origin of seismicity within the production region in relationship to production and injection processes is unclear. The second of these algorithms involves the derivation of a design criterion, which we apply to inform the expansion of the existing seismic monitoring programme at Kawerau geothermal field; we also apply an early version to the short-term/rapid-response network design following the M7.1 September 2010 Darfield earthquake. Unlike previous seismic network design algorithms, the new algorithm incorporates methods for the realistic representation of 3D velocity structures and attenuation models for both P and S travel times, a surface noise model, and the ability to apply complex weighting functions to the earthquake set. The results demonstrate the utility of this algorithm in even simplistic cases, and show how each new parameter incorporated into the design model affects the optimal network design obtained, identifying the need for accurate input data to provide optimal results.</p>

Microseismicity Associated with Actively Exploited Geothermal Systems: Earthquake Detection and Probabilistic Location at Rotokawa and Statistical Seismic Network Design at Kawerau

<p>Geothermal power has progressively been recognised as an important energy resource due to the depletion of old power sources, and as a more environmentally aware population pushes for an increase in renewable energy sources. Monitoring microseismicity occurring in active geothermal systems is one means of both characterising the system’s fault architecture and characterising fluid/rock interaction in response to production. This study focuses on better understanding seismicity in two active geothermal fields, through the development and implementation of two different algorithms: an automated microearthquake detection algorithm using a matched filter technique (improving earthquake detection), and an optimal seismic network design algorithm (improving earthquake location). Both algorithms have been implemented in codes that are easily adaptable to other data sets. The first of these algorithms has been applied to five months of continuous seismic waveform data spanning a fluid injection operation in the Rotokawa geothermal field. The cross-correlation of 14 high-quality master events with the continuous seismic data yields 2461 newly detected earthquakes spanning the magnitude range M=-0.4 to M=2.6 with a mean magnitude of M=0.47. The earthquakes detected with each master event exhibit high waveform similarity over approximately three orders of magnitude, and appear to follow a Gutenberg-Richter power law with a catalogue completeness down to M~ 0. Hypocentres for these detected events computed using the probabilistic earthquake location algorithm NonLinLoc reveal the dominant locus of seismicity to lie between 1.0–2.5 km depth, a location consistent with that of the Rotokawa Andesite which forms the Rotokawa reservoir. Focal mechanism solutions for the master events are predominantly normal, with half displaying a large strike-slip component, and the stress parameters obtained for this suite of focal mechanisms imply a northeast–southwest oriented maximum horizontal stress: both of these results are consistent with the extensional regime of the TVZ. Seismicity occurring within a 300 m horizontal radius of the injection well’s feed-zones, and extending to 5 km depth, initially exhibits a correlation with injection flow rates with a ~ 2 day lag, and seismicity rates decrease ~ 10 weeks after injection. We surmise that seismicity within the injection region and close to the injection well is likely to be injection-induced, with one portion of the injectate returning to the production region, while the other either migrates southeastward out of the field or remains within the injection region; the origin of seismicity within the production region in relationship to production and injection processes is unclear. The second of these algorithms involves the derivation of a design criterion, which we apply to inform the expansion of the existing seismic monitoring programme at Kawerau geothermal field; we also apply an early version to the short-term/rapid-response network design following the M7.1 September 2010 Darfield earthquake. Unlike previous seismic network design algorithms, the new algorithm incorporates methods for the realistic representation of 3D velocity structures and attenuation models for both P and S travel times, a surface noise model, and the ability to apply complex weighting functions to the earthquake set. The results demonstrate the utility of this algorithm in even simplistic cases, and show how each new parameter incorporated into the design model affects the optimal network design obtained, identifying the need for accurate input data to provide optimal results.</p>

Automatic inspection and analysis of digital waveform images by means of convolutional neural networks

Analyzing seismic data to get information about earthquakes has always been a major task for seismologists and, more in general, for geophysicists. Recently, thanks to the technological development of observation systems, more and more data are available to perform such tasks. However, this data “grow up” makes “human possibility” of data processing more complex in terms of required efforts and time demanding. That is why new technological approaches such as artificial intelligence are becoming very popular and more and more exploited. In this paper, we explore the possibility of interpreting seismic waveform segments by means of pre-trained deep learning. More specifically, we apply convolutional networks to seismological waveforms recorded at local or regional distances without any pre-elaboration or filtering. We show that such an approach can be very successful in determining if an earthquake is “included” in the seismic wave image and in estimating the distance between the earthquake epicenter and the recording station.

Site characterizations of BMKG seismic network, Indonesia, based on HVSR analysis

In the last few years BMKG (Agency for Meteorology Climatology and Geophysics) has increased the number of seismic stations significantly. Until mid-2020, when this study was conducted, the number of BMKG stations has reached 339 units. Development of the network is aimed to improve the data quality, speed of earthquake data processing and information dissemination to the public, accuracy of the hypocenter, as well as the magnitude. The objective of this study is to identify the site characteristics of the network. We analysed the noise recorded at BMKG stations throughout Indonesia using the HVSR (Horizontal to Vertical Spectral Ratio) method. The results of HVSR analysis were used to classify the site conditions of each station. We got the number of stations with the classification of hard rock, rock, hard soil, medium soil, and soft soil, 47, 57, 52, 30, and 84, respectively. This site condition represents the stations characteristics and affects the quality of seismic waveform data.

Application of P-Wave Moment Magnitude (Mwp) and Rupture Time Duration (Tdur) to Analyze A Potential Tsunami Earthquake in Sumatra

Tsunami warning is one of many important reports to save lives and reduce the damage for local peoples. A moment magnitude of P-wave (Mwp) and the rupture time duration (Tdur) can be used as the quickly parameters to diseminate the tsunami warning. In this paper, we analyze the seismic waveform from global network to get Mwp and Tdur of South-West Coast of Sumatera earthquake. Mwp was calculated using automatic and manual phase picking of P phase. The results of this study show a well-analyzed relationship between P wave from automatic and manual picking, Mwp and time duration, respectively. The result also give an encouraging studies for the early warning system that will be set up in the future in the region.

Real-Time Prediction of Impending Ground Shaking: Review of Wavefield-Based (Ground-Motion-Based) Method for Earthquake Early Warning

Earthquake early warning (EEW) systems aim to provide advance warning of impending ground shaking, and the technique used for real-time prediction of shaking is a crucial element of EEW systems. Many EEW systems are designed to predict the strength of seismic ground motions (peak ground acceleration, peak ground velocity, or seismic intensity) based on rapidly estimated source parameters (the source-based method), such as hypocentral location, origin time, magnitude, and extent of fault rupture. Recently, however, the wavefield-based (or ground-motion-based) method has been developed to predict future ground motions based directly on the current wavefield, i.e., ground motions monitored in real-time at neighboring sites, skipping the process of estimation of the source parameters. The wavefield-based method works well even for large earthquakes with long duration and huge rupture extents, highly energetic earthquakes that deviate from standard empirical relations, and multiple simultaneous earthquakes, for which the conventional source-based method sometimes performs inadequately. The wavefield-based method also enables prediction of the ongoing seismic waveform itself using the physics of wave propagation, thus providing information on the duration, in addition to the strength of strong ground motion for various frequency bands. In this paper, I review recent developments of the wavefield-based method, from simple applications using relatively sparse observation networks to sophisticated data assimilation techniques exploiting dense networks.