Monday 11 April 2011, by briandx // Summary and context


This project focuses on the use of the seismic ambient noise to monitor slight changes of properties in the solid Earth. The implications are the detection of change of strain at depth with applications in different contexts. A major field of application is the monitoring of potentially dangerous structures like volcanoes or active fault zones. The project includes new methodological developments, massive processing of existing data and field experiments. Applications in regions where changes are induced by human activity are important both for the quantitative refinement of the method and for the important economic and social implications of these problems.

Background and motivations

The members of our group were the first to succeed in using the records of seismic ambient noise to extract deterministic signals that allowed us to image the Earth’s interior with a high resolution. The technique is founded on the first principles of physics. The correlation of a random seismic wavefield measured at two distant points contains the Earth’s response between these points (Green function). In other words, if one registers a random wavefield at two separate points, he can construct virtual seismograms that exhibit all the characteristics of the ones produced in an active experiment where a source is located at one of the points and recorded at the other. Whereas there are in theory strong requirements for this property to hold perfectly, in practice, its use is efficient in most cases because of the nature of the excitation of the seismic noise and because of the Earth’s heterogeneity which results in a multiplication of scattered waves.

JPEG - 287.3 kb
Figure 1. Examples of noise-based seismic tomographies.

The analysis of ambient noise is a domain which has experienced rapid growth since our first papers (Campillo and Paul, 2003; Shapiro and Campillo, 2004; Shapiro et al.; 2005). The novel technique that we have proposed has been used by various groups to produce high-resolution images of the Earth’s crust in various regions in the world. The main advantages of this technique are that it provides new data for imaging, which are independent of the occurrence of earthquakes and measured along possibly short paths. Figure 1 presents two examples of investigation of geological structures.

More recently, we analyzed the stability and the accuracy of travel-time measurements obtained from correlations of seismic noise and discovered that they are remarkably robust. We therefore investigated the possibility of using repeated noise-based measurements as a monitoring tool. We found that we are able to detect relative temporal changes of the velocity smaller than 10-3. We applied this idea to the forecasting of volcanic eruptions and discovered at Piton de la Fournaise that a clearly detectable decrease in seismic velocity occurred a few days before eruptions (Brenguier et al., 2008). In addition to these precursory phenomena, the time history of velocity change exhibits slow fluctuations that may help us reveal information about long-term evolution of the volcano.

With better seismic networks and methodological developments, we will be able to further improve the accuracy of noise-based measurements, opening a new class of phenomena located at depth to continuous monitoring with seismological methods. Our main goal is to monitor changes associated with deep processes such as magma supply, tectonic loading, or co- or post-seismic deformation. The implications would be considerable for the forecast of volcanic eruptions, the monitoring of landslides and the search for co-seismic and possible precursory mechanical changes associated with earthquakes. Preliminary results indicate that velocity changes detected in the vicinity of fault zones are associated with tectonic processes. Velocity variations in the Parkfield area measured over seven years with a technique similar to the one used on the volcano (Figure 2) exhibit a remarkable correlation with the occurrence of earthquake and post-seismic relaxation. Moreover, the detected velocity changes are clearly correlated with the activity of non-volcanic tremors that occur at depth, likely below the seismogenic layer. This strongly suggests that the noise-based measurements allow us to detect variation in mechanical conditions at depth.

JPEG - 99.4 kb
Figure 2. Results of the study in Parkfield.
The grey curve denotes the relative change of seismic velocity deduced from the analysis of ambient noise. The green area represents the rate of activity of non-volcanic tremors (courtesy of R. Nadeau, UC Berkeley). Vertical lines correspond to the occurrence of two strong earthquakes. The red line is the post seismic relaxation function deduced from GPS following the 2004 Parkfield event.

Proposed research

Improving accuracy of noise-based measurements

For the dream of using noise to monitor the solid Earth to become reality, the key issue is to improve the accuracy of the measurements. We have to solve problems of two different natures. First, while a perfect Earth response reconstruction is expected for a perfectly random and stationary noise, the actual conditions are different and the real seismic noise field is far from being fully understood. One of the goals of our project is to use deterministic models of seismic noise deduced from oceanographic models of waves to correct for known inhomogeneous distribution of noise sources. We expect that precise wave action models with local directional spectra will help us evaluate the seismic excitation and simulate the seismic noise propagation from the source region to a specific point in a realistic Earth model. This information will help correcting the velocity measurements for the uneven distribution of energy flow. To create the excitation models, we will associate expertise in modeling oceanic waves and meteorology and in the domain of wave interaction and coupling with the solid Earth. At the same time, we will use actual seismic records to detect the apparent origin of the noise and to calibrate the predictions derived from oceanic wave action models. Another type of problem to be solved to improve accuracy of measurements is related to the processing of the seismic noise records. As a first step we propose to use refined pre-stack data adaptive filtering to select stable contributions. As a second step we propose to use the records from a network instead of a single pair of stations. With a network, we can use inter-station noise correlations as equivalents to seismograms produced by sources acting at station locations and correlate their ’coda’ parts. Preliminary results show that with this procedure, the reconstruction of the Green function is much more stable and less dependent on the properties of the seismic noise.

Interpretation of noise-based measurements

Standard approaches for travel-time measurements and inversion used in seismic imaging are not accurate enough to detect tiny changes in seismic speeds associated with transient deformations at depth. Our first results indicate that using the ’coda’ part of reconstructed Green functions allows us to dramatically increase the accuracy of repetitive differential measurements. Therefore, we plan to pursue theoretical and methodological research in order to develop this new type of measurements and their inversion for the 3D structure. We will perform numerical tests and laboratory experiments to analyze the possible resolution of this type of measurements and use our expertise in the radiative transfer theory to draw 3D maps of density of probability for the scattered energy at a given time and use them as a form of sensitivity kernels. We will systematically compare our results with seismicity, geodetic measurements of deformation, meteorological and hydrological observations, and collaborate with our colleagues specializing in studies of mechanical behavior of rocks to understand the relation between the observed changes in seismic speeds and the variation of mechanical conditions at depth.

Applications for studies of transient processes in the Earth’s crust

Our main goal is to apply the proposed noise-based methods to study the transient processes related to volcanic and tectonic activity through continuous measuring of mechanical changes in deep parts of the Earth. Our main research targets will be active regions and objects such as volcanoes, active faults, and subduction zones where transient deformation and associated phenomena have been detected recently. First, we will broadly use the continuous records provided by modern seismological networks. For studies of volcanoes, we will continue to process the data from La Réunion volcanological observatory and we will also look for the data available in other volcanic areas such as Japan, US Northwest, etc. We will continue our research in the Parkfield area and extend this study to a much broader area in southern and central California where one of the best broadband seismic networks has been continuously operating for several years. For studies of subduction zones, we will work with the networks operating in US Pacific Northwest and Japan. A particular focus of our studies will be the Mexican subduction zone, which is an almost ideal natural laboratory to study transient inter-seismic deformation because of its flat geometry and because of the very large size and duration of transient deformation events occurring in this region. In addition to work with existing data, we plan to carry out several dedicated seismological field experiments on the Piton de la Fournaise volcano, in Mexico and, as smaller scale experiments, in tectonically active regions in Europe.

Applications in controlled environments

In the absence of in situ control of the mechanical state at depth, it is necessary to validate our approach with controlled change at a scale somehow similar to that of geological objects like faults or volcanoes. There are several domains with potential applications for which the physical changes are controlled, including industrial applications such as hydrocarbon extraction or C02 storage. We plan to use passive records to correlate changes in seismic speeds with the known mechanical history of the rocks at depth. We have access to such data sets, which can be provided on a free basis by an oil company. At a smaller scale, we will monitor the surface of a tunnel during the excavation process to detect the evolution and the distribution of damage. At the laboratory scale, we are planning our own experiments involving changes in the medium produced by changes in strain and/or temperature.

Expected results and social and economical implications of the project

The seismic noise-based methods that we propose to develop would be combined with satellite-based geodetic observations to significantly improve our capability to study complex time-dependent and space-variable behavior underlying solid Earth deformation in response to tectonic and volcanic forcing and to environmental changes. First of all, we expect that these new developments will have applications in reducing human and economic losses due to natural hazards. We will continue to investigate the changes of velocity associated with the onset of volcanic eruptions to identify a systematic and reliable precursory signal. The next step will be to investigate the behavior of explosive volcanoes. Although there can be long intervals between eruptions, this type of volcano represents a major threat for large populations. Improving our forecasting capability is therefore very important. It requires monitoring over periods of time that are well beyond the duration of this proposal. It is nevertheless important to show the feasibility of this monitoring and to encourage the deployment of continuous recording systems when such a volcano enters an apparently critical stage.

No one can claim that efficient earthquake prediction is close at hand, and the very existence of a precursory phase with a duration sufficient for an alert to be issued is not guaranteed. Nevertheless, new discoveries have led to a revival of interest in the subject. ‘Slow’ transient deformation revealed by geodesists, as silent earthquakes, and non-volcanic tremors occurring at a depth where seismic activity was considered as non-existent, are discoveries that open new questions and new hopes. Our preliminary results (Figure 2) indicate that noise-based measurements are accurate enough to detect temporal changes of a living Earth. Correlations between the observed seismic speed changes and the occurrence of deep tremors suggest that the noise-based measurements are able to shed a light on mechanical processes at depth. This exciting field is open for new investigations, and could bring new elements to the discussion on the seismic cycle and eventually to the problem of earthquake prediction.

With the environmental impact of industrial activities becoming a growing subject of concern, a new monitoring capability is an important new tool to guarantee the security of underground activities. Extraction of oil and gas relies on massive injection and pumping that result in important variations of stress at depth. Monitoring has become an important issue for the industry, and it is even a growing concern with the likely development of new operations like CO2 storage or steam injection for tar extraction. Another domain in which monitoring is important is the safety of deep repositories of highly toxic wastes for which the long-term evolution of the surrounding rocks must by controlled.

Another important expected contribution is the training of young scientists and specialists. The PI, Michel Campillo, has a long record of teaching and supervising responsibilities, with many of his former students and postdocs now working as researchers and leaders in the academic world and in industry. Both the leading (UJF, LGIT) and the partner (IPGP) organizations are among the best French educational institutions with very strong undergraduate and graduate programs for fundamental and applied research in Earth sciences. UJF is the French partner of the Erasmus Mundus MEEES under the responsibility of LGIT researchers. The MEEES program is an Erasmus Mundus Masters Course, designed to provide higher-level education in the field of natural hazards. Both LGIT and IPGP are strongly involved with a European training network on Seismic wave Propagation and Imaging in Complex media (SPICE). The results of the proposed research will thus be transmitted very quickly to the new generation of young scientists and specialists who will be able to apply new knowledge and technologies in academic and industrial environments.

Research team and environment

The research group will be led by Michel Campillo, professor of Geophysics at Université Joseph Fourier (UJF), and leader of the team ‘Waves and Structures’ of Laboratoire de Géophysique Interne et Tectonophysique’ (LGIT). Two other LGIT permanent researchers are associated with the project: Philippe Roux and Eric Larose, both from the Centre National de la Recherche Scientifique (CNRS). The partner organization (Institut de Physique du Globe de Paris, IPGP) will be represented by Nikolai Shapiro, who is a senior researcher with CNRS and head of the IPGP seismology department, and by Florent Brenguier who is a researcher with the volcanological observatories. Four graduate students and two postdocs are now directly involved in research related to the present proposal. We are planning to maintain throughout the proposal this research team composed of three confirmed scientists (MC, PR, and NS), two very active young researchers (EL and FB), and a group of 5-6 graduate students and postdocs. An important part of the requested funding will be earmarked to maintain the size of this research group and to provide the participants with the necessary resources.

All permanent researchers involved have been actively participating in the recent development of the imaging and monitoring methods based on random seismic wavefields as can be seen from their publication record. Overall, they combine a broad range of expertise including seismology, the physics of wave propagation, and laboratory acoustics. Collaborations with colleagues in LGIT and IPGP and at international level will provide necessary links with tectonics, geodesy, volcanology, rock physics and physical oceanography.

Also, the leading and the partner organizations provide an important research infrastructure. The LGIT acoustical laboratory will be used for the planned experiments in controlled environment. Also, LGIT plays an important role in managing the French pool of seismic mobile instruments, providing us with the expertise and infrastructure necessary for carrying out the seismic field experiments. IPGP is operating the French volcanological observatories in the French Antilles and in La Réunion Island where we are planning to perform most of research related to development of the volcano monitoring methods. IPGP will also contribute its existing infrastructure and expertise in numerical computation and massive data processing.