Periodic Reporting for period 4 - F-IMAGE (Seismic Functional Imaging of the Brittle Crust)
Periodo di rendicontazione: 2022-04-01 al 2023-03-31
Earthquakes occur suddenly and without any unambiguous precursory signals identified to date. Our inability to develop prior warnings of earthquakes in the short term in purely natural or anthropically modified contexts, or to develop technologies that are ‘earthquake free’, is due to our poor understanding of the processes at depth that accompany earthquake occurrence. F-IMAGE aimed to provide an important contribution to this understanding with wide implications for seismic risk assessment and mitigation, and for industrial activities associated with induced seismicity, such as geothermal or mining.
The F-IMAGE project studies the rock around faults, and the mechanical behavior of the faults and their interactions with the rock around them. The methods of analysis to enable functional imaging of the Earth's crust need to be improved, or changed. Our developments rely heavily on continuous ambient seismic records, which are widely available and globally under-exploited. We characterize the fault zone in terms not only of classical seismic velocities, but also of its scattering strength related to complexity and damage. We aim to describe how these structures evolve in time, or to be more precise, to monitor some observables that can be related to the physical processes at work.
We develop and improve methods to measure new observables such as: temporal evolution of elastic properties in the vicinity of faults; changes in scattering properties as indicative of damage to the rock; time-dependent classification of the various components of the continuous motion of the Earth surface (e.g. micro-earthquakes, tremors, noise of different origins); statistical behavior of the seismic activity in relation to other observations, and primarily geodetic measurements. Our guess is that with such new data in hand, obtained through the more recent progress in seismology and geodesy, we open new fields of research that are associated with the new observables and the weak signals that have been discarded in the past.
These new techniques have been applied to high-risk areas and have shown their potential for the imaging of fluid movements, direct measurement of slow slip in fault roots, localization of shear in the mantle or discovery of transient deformation at large depth. The methodological developments of F-IMAGE are useful well beyond this present project with already applications in water resource management, volcanology or glaciology.
The F-IMAGE project studies the rock around faults, and the mechanical behavior of the faults and their interactions with the rock around them. The methods of analysis to enable functional imaging of the Earth's crust need to be improved, or changed. Our developments rely heavily on continuous ambient seismic records, which are widely available and globally under-exploited. We characterize the fault zone in terms not only of classical seismic velocities, but also of its scattering strength related to complexity and damage. We aim to describe how these structures evolve in time, or to be more precise, to monitor some observables that can be related to the physical processes at work.
We develop and improve methods to measure new observables such as: temporal evolution of elastic properties in the vicinity of faults; changes in scattering properties as indicative of damage to the rock; time-dependent classification of the various components of the continuous motion of the Earth surface (e.g. micro-earthquakes, tremors, noise of different origins); statistical behavior of the seismic activity in relation to other observations, and primarily geodetic measurements. Our guess is that with such new data in hand, obtained through the more recent progress in seismology and geodesy, we open new fields of research that are associated with the new observables and the weak signals that have been discarded in the past.
These new techniques have been applied to high-risk areas and have shown their potential for the imaging of fluid movements, direct measurement of slow slip in fault roots, localization of shear in the mantle or discovery of transient deformation at large depth. The methodological developments of F-IMAGE are useful well beyond this present project with already applications in water resource management, volcanology or glaciology.
For each of the specific topics of F-IMAGE, our work consists of both theoretical and methodological advances and applications to actual data. The specificity of our study is to fully exploit widely available long continuous seismic records of ambient vibrations.
We have developed novel approaches to image and characterize the medium around a main fault according to its scattering properties. We used either coda energy analysis or a passive matrix technique, which has required a new strategy of corrections of the aberrations induced by the strong velocity variations existing in the Earth. We applied these approaches to the shallow part of the San Jacinto Fault in California, at a crustal scale to the North Anatolian Fault Zone.
We have improved our techniques for monitoring the rock properties on their different aspects. A key issue is to develop our ability to precisely locate the temporal changes that we detect with coda waves. We have shown the important effect of a non-uniform scattering to the sensitivity of coda wave to a local change. We have developed a theory including the conversions between body-wave and surface-wave that accounts for the sensitivity obtained in fully numerical studies. We have studied how seismic velocity (or scattering strength) in the entire crust responds to several processes, including the earthquake related fluids motions at depths and the water load fluctuation related with climate and human activities.
The multi-scale representation of seismic signals associated with machine learning tools allows for unsupervised detection of the evolutions of seismic activity and ambient noise. Continuous descriptors/features can be related to physical processes at work. We improved array-matched filter detection by introducing a stage of machine learning, and we built high quality catalogs that can be used to detect faint geodetic signals, such as repeated slow-slip events that occur in the roots of active faults.
We have also performed laboratory experiments and numerical simulation with non-linear cohesive granular media to mimic the velocity changes occurring in a damage material subject to deformation in nature. We have shown that the large stress variations at small scale result in the extreme sensitivity of macroscopic seismic speed. We introduced fault complexity in our models, including a time-dependent damage in a viscoelastic body for the long-term evolution of a fault.
We have developed novel approaches to image and characterize the medium around a main fault according to its scattering properties. We used either coda energy analysis or a passive matrix technique, which has required a new strategy of corrections of the aberrations induced by the strong velocity variations existing in the Earth. We applied these approaches to the shallow part of the San Jacinto Fault in California, at a crustal scale to the North Anatolian Fault Zone.
We have improved our techniques for monitoring the rock properties on their different aspects. A key issue is to develop our ability to precisely locate the temporal changes that we detect with coda waves. We have shown the important effect of a non-uniform scattering to the sensitivity of coda wave to a local change. We have developed a theory including the conversions between body-wave and surface-wave that accounts for the sensitivity obtained in fully numerical studies. We have studied how seismic velocity (or scattering strength) in the entire crust responds to several processes, including the earthquake related fluids motions at depths and the water load fluctuation related with climate and human activities.
The multi-scale representation of seismic signals associated with machine learning tools allows for unsupervised detection of the evolutions of seismic activity and ambient noise. Continuous descriptors/features can be related to physical processes at work. We improved array-matched filter detection by introducing a stage of machine learning, and we built high quality catalogs that can be used to detect faint geodetic signals, such as repeated slow-slip events that occur in the roots of active faults.
We have also performed laboratory experiments and numerical simulation with non-linear cohesive granular media to mimic the velocity changes occurring in a damage material subject to deformation in nature. We have shown that the large stress variations at small scale result in the extreme sensitivity of macroscopic seismic speed. We introduced fault complexity in our models, including a time-dependent damage in a viscoelastic body for the long-term evolution of a fault.
The goals of our proposal were ambitious and indicated that we wanted to overcome a series of theoretical and technical difficulties. Our new techniques for imaging fault zone heterogeneity have allowed us to individualize the different fault strands apparent at the surface and to visualize their deep structures down to the mantle with important implications for the mechanics of fault systems. This has been made possible by the development of new approaches to overcome the problem of aberrations due to surface velocity variations.
We have measured temporal changes in the mechanical properties of the whole crust related to the occurrence of earthquakes or environmental forcings. In particular, we observed the transfer of fluids from the base of the crust to the surface in a few months in Japan. We have advanced the field methodologically by considering the coupling between body and surface waves and the non-uniformity of scattering.
We have developed a new method of unsupervised signal analysis based on a multi-scale representation. It allowed to detect the evolution of different categories of seismic signals and to extract continuous features associated with physical processes. The improvement of array match-filter detection methods with the introduction of deep learning tools allows to establish high quality catalogs without a priori. We have used this type of catalog to finely analyze geodetic positioning series, which allowed the first direct observation of transient motions in the root of the San Andreas Fault.
We reproduced features of seismological observations in laboratory experiments with nonlinear cohesive granular materials taken as analogues of highly damaged fault zones. Numerical simulation allowed us to show how macroscopic velocity behavior was the result of stress heterogeneities at the microscopic scale.
Our time-dependent imaging methods have applications in other important fields. This is the case with the monitoring of water resources with the permanent seismological networks already in place or the monitoring of ice layers, two major applications with respect to climate change.
New analytical methods offer new insights and make room for serendipity. We have thus highlighted a new phenomenon: a transient change in the slab pull on the Kamchatka subduction that challenges fundamental assumptions in the study of the seismic cycle and the mechanical relationships between megathrusts and deep earthquakes.
We have measured temporal changes in the mechanical properties of the whole crust related to the occurrence of earthquakes or environmental forcings. In particular, we observed the transfer of fluids from the base of the crust to the surface in a few months in Japan. We have advanced the field methodologically by considering the coupling between body and surface waves and the non-uniformity of scattering.
We have developed a new method of unsupervised signal analysis based on a multi-scale representation. It allowed to detect the evolution of different categories of seismic signals and to extract continuous features associated with physical processes. The improvement of array match-filter detection methods with the introduction of deep learning tools allows to establish high quality catalogs without a priori. We have used this type of catalog to finely analyze geodetic positioning series, which allowed the first direct observation of transient motions in the root of the San Andreas Fault.
We reproduced features of seismological observations in laboratory experiments with nonlinear cohesive granular materials taken as analogues of highly damaged fault zones. Numerical simulation allowed us to show how macroscopic velocity behavior was the result of stress heterogeneities at the microscopic scale.
Our time-dependent imaging methods have applications in other important fields. This is the case with the monitoring of water resources with the permanent seismological networks already in place or the monitoring of ice layers, two major applications with respect to climate change.
New analytical methods offer new insights and make room for serendipity. We have thus highlighted a new phenomenon: a transient change in the slab pull on the Kamchatka subduction that challenges fundamental assumptions in the study of the seismic cycle and the mechanical relationships between megathrusts and deep earthquakes.