Heterogeneous cloud-supercomputing framework for daily seismic noise source inversion

2020 ◽  
Author(s):  
Alexey Gokhberg ◽  
Laura Ermert ◽  
Jonas Igel ◽  
Andreas Fichtner
Keyword(s):  
Real Time ◽  
Seismic Noise ◽  
Cross Correlation ◽  
Noise Source ◽  
Workflow Management ◽  
Source Distribution ◽  
Ambient Seismic Noise ◽  
Source Inversion ◽  
Finite Frequency ◽  
Noise Sources

<p>The study of ambient seismic noise sources and their time- and space-dependent distribution is becoming a crucial component of the real-time monitoring of various geosystems, including active fault zones and volcanoes, as well as geothermal and hydrocarbon reservoirs. In this context, we have previously implemented a combined cloud - HPC infrastructure for production of ambient source maps with high temporal resolution. It covers the entire European continent and the North Atlantic, and is based on seismic data provided by the ORFEUS infrastructure. The solution is based on the Application-as-a-Service concept and includes (1) acquisition of data from distributed ORFEUS data archives, (2) noise source mapping, (3) workflow management, and (4) front-end Web interface to end users.</p><p>We present the new results of this ongoing project conducted with support of the Swiss National Supercomputing Centre (CSCS). Our recent goal has been transitioning from mapping the seismic noise sources towards modeling them based on our new method for near real-time finite-frequency ambient seismic noise source inversion. To invert for the power spectral density of the noise source distribution of the secondary microseisms we efficiently forward model global cross-correlation wavefields for any noise distribution. Subsequently, a gradient-based iterative inversion method employing finite-frequency sensitivity kernels is implemented to reduce the misfit between synthetic and observed cross correlations.</p><p>During this research we encountered substantial challenges related to the large data volumes and high computational complexity of involved algorithms. We handle these problems by using the CSCS massively parallel heterogeneous supercomputer "Piz Daint". We also apply various specialized numeric techniques which include: (1) using precomputed Green's functions databases generated offline with Axisem and efficiently extracted with Instaseis package and (2) our previously developed high performance package for massive cross correlation of seismograms using GPU accelerators. Furthermore, due to the inherent restrictions of supercomputers, some crucial components of the processing pipeline including the data acquisition and workflow management are deployed on the OpenStack cloud environment. The resulting solution combines the specific advantages of the supercomputer and cloud platforms thus providing a viable distributed platform for the large-scale modeling of seismic noise sources.</p>

Rapid Global Finite-Frequency Ambient Noise Source Inversion

2020 ◽  
Author(s):  
Jonas Igel ◽  
Laura Ermert ◽  
Andreas Fichtner
Keyword(s):  
Ambient Noise ◽  
Noise Source ◽  
Computational Cost ◽  
User Interaction ◽  
Real Data ◽  
Source Distribution ◽  
Source Inversion ◽  
Finite Frequency ◽  
Noise Sources ◽  
Cross Correlations

<p>Common assumptions in ambient noise seismology such as Green’s function retrieval and equipartitioned wavefields are often not met in the Earth. Full waveform ambient noise tomography methods are free of such assumptions, as they implement knowledge of the time- and space-dependent ambient noise source distribution, whilst also taking finite-frequency effects into account. Such methods would greatly simplify near real-time monitoring of the sub-surface. Additionally, the distribution of the secondary microseisms could act as a new observable of the ocean state since its mechanism is well understood (e.g. Ardhuin et al., 2011).</p><p>To efficiently forward-model global noise cross-correlations we implement (1) pre-computed high-frequency wavefields obtained using, for example, AxiSEM (Nissen-Meyer et al., 2014), and (2) spatially variable grids, both of which greatly reduce the computational cost. Global cross-correlations for any source distribution can be computed within a few seconds in the microseismic frequency range (up to 0.2 Hz). Similarly, we can compute the finite-frequency sensitivity kernels which are then used to perform a gradient-based iterative inversion of the power-spectral density of the noise source distribution. We take a windowed logarithmic energy ratio of the causal and acausal branches of the cross-correlations as measurement, which is largely insensitive to unknown 3D Earth structures.</p><p>Due to its parallelisation on a cluster, our inversion tool is able to rapidly invert for the global microseismic noise source distribution with minimal required user interaction. Synthetic and real data inversions show promising results for noise sources in the North Atlantic with the structure and spatial distribution resolved at scales of a few hundred kilometres. Finally, daily noise sources maps could be computed by combining our inversion tool with a daily data download and processing toolkit.</p>

scholarly journals Imaging New Zealand's Crustal Structure Using Ambient Seismic Noise Recordings from Permanent and Temporary Instruments

2021 ◽  
Author(s):  
◽  
Yannik Behr
Keyword(s):  
Seismic Noise ◽  
Cross Correlation ◽  
Velocity Structure ◽  
Shear Velocity ◽  
Ambient Seismic Noise ◽  
Uppermost Mantle ◽  
Noise Sources ◽  
Noise Field ◽  
Rayleigh And Love Waves ◽  
Noise Cross Correlation

<p>We use ambient seismic noise to image the crust and uppermost mantle, and to determine the spatiotemporal characteristics of the noise field itself, and examine the way in which those characteristics may influence imaging results. Surface wave information extracted from ambient seismic noise using cross-correlation methods significantly enhances our knowledge of the crustal and uppermost mantle shear-velocity structure of New Zealand. We assemble a large dataset of three-component broadband continuous seismic data from temporary and permanent seismic stations, increasing the achievable resolution of surface wave velocity maps in comparison to a previous study. Three-component data enables us to examine both Rayleigh and Love waves using noise cross-correlation functions. Employing a Monte Carlo inversion method, we invert Rayleigh and Love wave phase and group velocity dispersion curves separately for spatially averaged isotropic shear velocity models beneath the Northland Peninsula. The results yield first-order radial anisotropy estimates of 2% in the upper crust and up to 15% in the lower crust, and estimates of Moho depth and uppermost mantle velocity compatible with previous studies. We also construct a high-resolution, pseudo-3D image of the shear-velocity distribution in the crust and uppermost mantle beneath the central North Island using Rayleigh and Love waves. We document, for the first time, the lateral extent of low shear-velocity zones in the upper and mid-crust beneath the highly active Taupo Volcanic Zone, which have been reported previously based on spatially confined 1D shear-velocity profiles. Attributing these low shear-velocities to the presence of partial melt, we use an empirical relation to estimate an average percentage of partial melt of < 4:2% in the upper and middle crust. Analysis of the ambient seismic noise field in the North Island using plane wave beamforming and slant stacking indicates that higher mode Rayleigh waves can be detected, in addition to the fundamental mode. The azimuthal distributions of seismic noise sources inferred from beamforming are compatible with high near-coastal ocean wave heights in the period band of the secondary microseism (~7 s). Averaged over 130 days, the distribution of seismic noise sources is azimuthally homogeneous, indicating that the seismic noise field is well-suited to noise cross-correlation studies. This is underpinned by the good agreement of our results with those from previous studies. The effective homogeneity of the seismic noise field and the large dataset of noise cross-correlation functions we here compiled, provide the cornerstone for future studies of ambient seismic noise and crustal shear velocity structure in New Zealand.</p>

scholarly journals Imaging New Zealand's Crustal Structure Using Ambient Seismic Noise Recordings from Permanent and Temporary Instruments

2021 ◽  
Author(s):  
◽  
Yannik Behr
Keyword(s):  
Seismic Noise ◽  
Cross Correlation ◽  
Velocity Structure ◽  
Shear Velocity ◽  
Ambient Seismic Noise ◽  
Uppermost Mantle ◽  
Noise Sources ◽  
Noise Field ◽  
Rayleigh And Love Waves ◽  
Noise Cross Correlation

<p>We use ambient seismic noise to image the crust and uppermost mantle, and to determine the spatiotemporal characteristics of the noise field itself, and examine the way in which those characteristics may influence imaging results. Surface wave information extracted from ambient seismic noise using cross-correlation methods significantly enhances our knowledge of the crustal and uppermost mantle shear-velocity structure of New Zealand. We assemble a large dataset of three-component broadband continuous seismic data from temporary and permanent seismic stations, increasing the achievable resolution of surface wave velocity maps in comparison to a previous study. Three-component data enables us to examine both Rayleigh and Love waves using noise cross-correlation functions. Employing a Monte Carlo inversion method, we invert Rayleigh and Love wave phase and group velocity dispersion curves separately for spatially averaged isotropic shear velocity models beneath the Northland Peninsula. The results yield first-order radial anisotropy estimates of 2% in the upper crust and up to 15% in the lower crust, and estimates of Moho depth and uppermost mantle velocity compatible with previous studies. We also construct a high-resolution, pseudo-3D image of the shear-velocity distribution in the crust and uppermost mantle beneath the central North Island using Rayleigh and Love waves. We document, for the first time, the lateral extent of low shear-velocity zones in the upper and mid-crust beneath the highly active Taupo Volcanic Zone, which have been reported previously based on spatially confined 1D shear-velocity profiles. Attributing these low shear-velocities to the presence of partial melt, we use an empirical relation to estimate an average percentage of partial melt of < 4:2% in the upper and middle crust. Analysis of the ambient seismic noise field in the North Island using plane wave beamforming and slant stacking indicates that higher mode Rayleigh waves can be detected, in addition to the fundamental mode. The azimuthal distributions of seismic noise sources inferred from beamforming are compatible with high near-coastal ocean wave heights in the period band of the secondary microseism (~7 s). Averaged over 130 days, the distribution of seismic noise sources is azimuthally homogeneous, indicating that the seismic noise field is well-suited to noise cross-correlation studies. This is underpinned by the good agreement of our results with those from previous studies. The effective homogeneity of the seismic noise field and the large dataset of noise cross-correlation functions we here compiled, provide the cornerstone for future studies of ambient seismic noise and crustal shear velocity structure in New Zealand.</p>

Imaging noise sources: comparing and combining a matched-field processing technique with finite-frequency noise source inversions

2021 ◽  
Author(s):  
Jonas Igel ◽  
Daniel Bowden ◽  
Korbinian Sager ◽  
Andreas Fichtner
Keyword(s):  
Noise Source ◽  
Time Window ◽  
Waveform Inversion ◽  
Temporal Variations ◽  
Source Distribution ◽  
Frequency Noise ◽  
Finite Frequency ◽  
Noise Sources ◽  
Matched Field Processing ◽  
Full Waveform

&lt;p&gt;Imaging the spatio-temporal variations of ambient seismic noise sources can provide important information to improve near real-time monitoring and noise tomography. Various methods have been developed to tackle this problem. For example, Matched-Field Processing (MFP) offers an efficient data-driven approach by testing different noise source locations and subsequently correlating and stacking. A more rigorous approach is treating it as a finite-frequency full-waveform inversion problem. In contrast to the MFP technique, an inversion framework allows for the incorporation of prior information and subsequent iterative updates of the noise source distribution by numerically modelling correlations and source sensitivity kernels. Bowden et al. (2020) discuss the similarities between these two methods and how one can be derived from the other.&amp;#160;&lt;/p&gt;&lt;p&gt;We aim to compare and contrast the two methods using real data from a regional to a global scale to locate the secondary microseismic sources in the ocean. Igel et al. (2021, in prep) use a logarithmic energy ratio as measurement for the sensitivity kernels, which is chosen due to its robustness with respect to unknown 3D Earth structures. However, some disadvantages of this type of measurement are not considering absolute amplitudes and discarding information outside of the expected surface wave arrival time window. By combining the two methods and first using MFP to create an initial model for the inversion, we are able to steer the inversion in the right direction, allowing us to use a more elaborate full-waveform measurement in the inversion and hence increasing the resolution and quality of the final model.&amp;#160;&lt;/p&gt;&lt;p&gt;Results for noise source inversions in the ocean on a daily basis using the combination of the two methods will be presented. This work paves the way for publicly available, daily, multi-scale ambient noise source maps.&lt;/p&gt;

SeisNoise.jl: Ambient Seismic Noise Cross Correlation on the CPU and GPU in Julia

2020 ◽  
Vol 92 (1) ◽  
pp. 517-527
Author(s):  
Timothy Clements ◽  
Marine A. Denolle
Keyword(s):  
Seismic Noise ◽  
High Performance ◽  
Cross Correlation ◽  
Graphic Processing Unit ◽  
Ambient Seismic Noise ◽  
Processing Unit ◽  
Central Processing ◽  
And Performance ◽  
Noise Cross Correlation ◽  
Performance Computing

Abstract We introduce SeisNoise.jl, a library for high-performance ambient seismic noise cross correlation, written entirely in the computing language Julia. Julia is a new language, with syntax and a learning curve similar to MATLAB (see Data and Resources), R, or Python and performance close to Fortran or C. SeisNoise.jl is compatible with high-performance computing resources, using both the central processing unit and the graphic processing unit. SeisNoise.jl is a modular toolbox, giving researchers common tools and data structures to design custom ambient seismic cross-correlation workflows in Julia.

scholarly journals Passive monitoring of a sea dike during a tidal cycle using sea waves as a seismic noise source

2018 ◽  
Vol 214 (2) ◽  
pp. 1364-1378 ◽  
Author(s):  
Anaëlle Joubert ◽  
Mathieu Le Feuvre ◽  
Philippe Côte
Keyword(s):  
Seismic Noise ◽  
Tidal Cycle ◽  
Cross Correlation ◽  
Water Infiltration ◽  
Internal Erosion ◽  
Spectral Amplitude ◽  
Ambient Seismic Noise ◽  
Sea Waves ◽  
Reconstructed Surface ◽  
Sea Dike

SUMMARY Over the past decade, ambient seismic noise has been used successfully to monitor various geological objects with high accuracy. Recently, it has been shown that surface seismic waves propagating within a sea dike body can be retrieved from the cross-correlation of ambient seismic noise generated by sea waves. We use sea wave impacts to monitor the response of a sea dike during a tidal cycle using empirical Green’s functions. These are obtained either by cross-correlation or deconvolution, from signals recorded by sensors installed linearly on the crest of a dike. Our analysis is based on delay and spectral amplitude measurements performed on reconstructed surface waves propagating along the array. We show that localized variations of velocity and attenuation are correlated with changes in water level as a probable consequence of water infiltration inside the structure. Sea dike monitoring is of critical importance for safety and economic reasons, as internal erosion is generally only detected at late stages by visual observations. The method proposed here may provide a solution for detecting structural weaknesses, monitoring progressive internal erosion and delineating areas of interest for further geotechnical studies, in view to understand the erosion mechanisms involved.

scholarly journals Structure of the Suasselkä postglacial fault in northern Finland obtained by analysis of local events and ambient seismic noise

Solid Earth ◽  
2017 ◽  
Vol 8 (2) ◽  
pp. 531-544 ◽  
Author(s):  
Nikita Afonin ◽  
Elena Kozlovskaya ◽  
Ilmo Kukkonen ◽  
Keyword(s):  
Seismic Noise ◽  
Ambient Seismic Noise ◽  
Seismic Events ◽  
S Wave ◽  
Noise Sources ◽  
Surface Wave Dispersion ◽  
Velocity Models ◽  
Wave Velocities ◽  
Short Period ◽  
Single Station

Abstract. Understanding the inner structure of seismogenic faults and their ability to reactivate is particularly important in investigating the continental intraplate seismicity regime. In our study we address this problem using analysis of local seismic events and ambient seismic noise recorded by the temporary DAFNE array in the northern Fennoscandian Shield. The main purpose of the DAFNE/FINLAND passive seismic array experiment was to characterize the present-day seismicity of the Suasselkä postglacial fault (SPGF), which was proposed as one potential target for the DAFNE (Drilling Active Faults in Northern Europe) project. The DAFNE/FINLAND array comprised an area of about 20 to 100 km and consisted of eight short-period and four broadband three-component autonomous seismic stations installed in the close vicinity of the fault area. The array recorded continuous seismic data during September 2011–May 2013. Recordings of the array have being analysed in order to identify and locate natural earthquakes from the fault area and to discriminate them from the blasts in the Kittilä gold mine. As a result, we found a number of natural seismic events originating from the fault area, which proves that the fault is still seismically active. In order to study the inner structure of the SPGF we use cross-correlation of ambient seismic noise recorded by the array. Analysis of azimuthal distribution of noise sources demonstrated that during the time interval under consideration the distribution of noise sources is close to the uniform one. The continuous data were processed in several steps including single-station data analysis, instrument response removal and time-domain stacking. The data were used to estimate empirical Green's functions between pairs of stations in the frequency band of 0.1–1 Hz and to calculate corresponding surface wave dispersion curves. The S-wave velocity models were obtained as a result of dispersion curve inversion. The results suggest that the area of the SPGF corresponds to a narrow region of low S-wave velocities surrounded by rocks with high S-wave velocities. We interpret this low-velocity region as a non-healed mechanically weak fault damage zone (FDZ) formed due to the last major earthquake that occurred after the last glaciation.

Incorporating spatial gradient observations into seismic noise interferometry

2021 ◽  
Author(s):  
Patrick Paitz ◽  
Korbinian Sager ◽  
Christian Boehm ◽  
Andreas Fichtner
Keyword(s):  
Noise Source ◽  
Source Distribution ◽  
Seismic Interferometry ◽  
Spatial Gradient ◽  
Source Inversion ◽  
Common Assumption ◽  
Wide Range ◽  
The Common ◽  
Arbitrary Noise ◽  
Public Datasets

&lt;p&gt;With an increasing availability of next-generation instruments in seismology such as Distributed Acoustic Sensing (DAS) interrogators and rotation sensors, as well as public datasets from these instruments, there is a demand for incorporating these new gradient observables into the workflows of seismic interferometry and noise source inversion.&lt;/p&gt;&lt;p&gt;Dropping the common assumption of Green&amp;#8217;s function retrieval, we derive a generalized formulation for seismic interferometry that can utilize not only displacement measurements but also spatial and temporal gradients thereof &amp;#8211; including velocity, strain and rotation.&lt;/p&gt;&lt;p&gt;Based on this formulation, we are able to simulate interferometric wavefields of displacement and gradient observations or arbitrary combinations of these observables, for heterogeneous visco-elastic media, and for arbitrary noise source distributions.&lt;/p&gt;&lt;p&gt;We demonstrate how to derive adjoint-based expressions for finite-frequency sensitivity kernels of the interferometric wavefields with respect to subsurface structure and noise source distributions, for a wide range of observed quantitates and combinations thereof. We provide numerical examples of such sensitivity kernels.&lt;/p&gt;&lt;p&gt;Especially in environments where the common assumption of a homogeneous noise source distribution is violated, our formulation enables correlation-wavefield based inversions, combining different seismic observables.&lt;/p&gt;&lt;p&gt;The discussed theoretical and numerical developments bring us one step closer to multi-observational full waveform ambient noise inversion, underlining the potential and possible impact of recent developments in seismic instrumentation to seismology across all scales.&lt;/p&gt;

Sign in / Sign up

Export Citation Format

Share Document