My research interests embrace different aspects of solid earth geophysics including, but not limited to: earthquake seismology, real-time seismology, early warning, seismic noise correlations, volcano seismology, surface-wave analysis for near-surface characterization. In my research, I explore various questions related to earthquake and volcano seismology leading to a better understanding of physical phenomena involved in tectonic and volcanic environments. The insights and techniques gained from such studies can be applied to reduce the impact of natural phenomena on our society (e.g., earthquakes, tsunamis, volcanic eruptions).
It is known that many large earthquakes involve significant rupture complexity, with the most well-known recent events being the Mw=7.1 1999 Hector Mine earthquake the Mw=8.1 2009 Samoa earthquake and the Mw=8.6 2012 Sumatra earthquake. Rupture branching onto multiple subfaults are frequently observed, for example, during the Mw=7.3 1992 Landers earthquake, during the Mw=7.8 2001 Kunlun earthquake and more recently during the Mw=7.2 2010 El Mayor-Cucapah earthquake. Such complex ruptures are generally difficult to characterize and several month are often necessary before having a reliable description of the source.
To account for the source complexity of large earthquakes, I developed a multiple-point-source moment tensor inversion based on long-period seismic waves. This method allows to retrieve the subevent moment tensors, locations (latitudes, longitudes, depths), time delays and half durations using a modified version of the Neighborhood Algorithm sampler (Sambridge, 1999). In Duputel et al. (2012, cf. publications), we validate this method for the Mw=8.6 2012 Sumatra earthquake. Our preferred two-point-source model for this earthquake is shown in the figure bellow as the green mechanisms an circles labeled I (first subevent) and II (second subevent).
Map of the 2012 Sumatra great earthquake region. The 11 April 2012 mainshock can be decomposed into two subevents separated by about 200 km (green mechanisms and circles labeled I and II). The W phase and Global CMT (GCMT) single-point-source solutions for the mainshock (inset green mechanisms), the W phase solutions for the 10 January foreshock (blue mechanism), for the Mw=8.2 aftershock (yellow mechanism) and for the 5.8 ≤ Mw ≤ 8.2 aftershocks (red mechanisms) are shown. Yellow circles indicate the earthquake epicenters and magnitudes from the National Earthquake Information Center (NEIC) catalog between 1 January 1973 and 10 April 2012. Red circles show the events since the Mw=8.6 11 April 2012 earthquake through May 2012. White arrows indicate the direction of motion of the Australian plate relative to the Indian plate at about 13 mm/yr.
Considerable effort has been made in the last two decades regarding the design and implementation of tools aimed at fast characterization of earthquake sources. But until recently, it still took several hours to determine the first order attributes of a major earthquake (e.g. Mw ≥ 7.5), even in a well instrumented region. A very good example is the Mw=9.2 2004 Sumatra-Andaman earthquake for which several days of discussion were necessary before having a concensus on the event magnitude. During my Ph.D. thesis, I developped a method to speedup the source analysis of such large events. This method is based on the W phase, a very-long-period phase (100 - 1000 s) starting at the same time as the P-wave. The W phase is conspicuous on broadband displacement records before the surface-waves arrivals and it can be interpreted as the superposition of the first overtones of the Earth normal modes at long period.
In Duputel et al. (2012, cf. publications), we show the robustness of this method at teleseismic distances by inverting for the centroid moment tensor of Mw ≥ 6.0 earthquakes reported globally between 1990 and 2010 (∼2500 events, cf. figure bellow for Mw ≥ 6.5 earthquakes). Several international collaborations have been established in the last three years for the real-time implemention of the method. The W phase algorithm is now used routinely in several warning centers like the Pacific Tsunami Warning Center (PTWC) or the United States Geological Survey (USGS) for the rapid source characterization of large earthquakes. During the recent Tohoku-oki earthquake, the W phase method provided a first Mw=9.0 estimate 20 min only after the origin time, while most other magnitude calculations severly underestimated the true event size.
One of the future improvements of the W phase inversion method is to speedup the source analysis using regional data. We recently pointed out the possibility to use F-NET Japanese stations within ∆ ≤ 12.0° to obtain a solution ∼6 min after the origin time. In Rivera et al. (2011, cf. publications), we show the possibility of using high-rate GPS data to decrease this delay to ∼5 min if the data is available in real-time. The algorithm is currently under implementation in several places like in Mexico, South-California and Japan. Another perspective related to the W phase algorithm is the possibility to rapidly characterize the source finiteness. However, to preserve the robustness of the inversion, we should extend the model with a minimum set of well chosen additional parameters on which we can have rapid and reliable estimates.
W-phase CMT solutions obtained for all Mw>=6.5 earthquakes occurring between 1990-01-01 and 2011-01-01.
Mar 2011 2011 Tohoku-oki earthquake: W-phase source inversion results are available online at http://eost.u-strasbg.fr/wphase/events/tohoku_oki_2011.
Oct 2010 A real-time implementation of the W-phase algorithm is currently under implementation at the Geophisical institute of UNAM for rapid CMT determinations of large eartquakes in Mexico.
May 2010 Recent developments lead to the use of the three components of displacement which can be very useful especially at regional scale. An example of regional application for chile and japan has been presented at the AGU 2010 Chapman conference in Chile. The poster is available from my publications page.
April 2010 The W-phase source inversion algorithm proved successful real-time operation for the recent Chilean event (Mw8.8). W-Phase solutions comming from different agencies have been presented at the 2010 SSA Annual Meeting. Posters and abstracts can be downloaded directly from my publications page.
Dec 2009 Recent advances in the algorithm has been presented at the AGU 2009 fall meeting. The posters are available from my publications page.
Oct 2009 A new version of the W-phase source inversion algorithm is now available and is currently under implementation in several warning centers : PTWC (Hawai'i), BATS (Taiwan), Mexican National Seismic Network.
Jul 2009 We developed a real-time W-phase source inversion in order to evaluate the robustness of three-component W-phase solutions for rapid earthquake response.
Earthquake source inversion is a widely used practice in seismology. Magnitudes, moment tensors, slip distributions are now routinely calculated and disseminated whenever an event occurs. The accuracy of such models depends on many aspects like the earthquake magnitude, the data coverage and the data quality (instrument response, isolation, timing, etc.). Here, like in any observational problem, the error estimation should be part of the solution. It is however very rare to find a source inversion algorithm which includes realistic error analyses, and the solutions are often given without any estimates of uncertainty.
In Duputel et al. (2012, cf. publications), we stress the importance of such estimation and explore different techniques aimed at achieving such analyses. We focus in particular on the linear-inverse problem of estimating the moment tensor components at a given source location and we assume that the initial probability densities can be modeled by Gaussian distributions. Formally, we can separate two sources of error which generally contribute to the model parameter uncertainties. The first source of uncertainty is the error introduced by the more or less imperfect data. The second source of uncertainty, often overlooked, is associated with modeling error or mismodeling. Among the different sources of mismodeling, we focus here on the modeling error associated with the mislocation of the centroid position. In source inversion problems, like in many other fields of geophysics, the data covariance (CD) is often considered as diagonal or even proportional to the identity matrix. In Duputel et al. (2012), we demonstrate the importance of using a more realistic form for CD.Taking into account more realistic data uncertainties allows us to improve error estimates on the source model parameters but also to improve the solution itself.
The earth is often affected by transient or permanent variation in its elastic properties. These perturbations will particularly affect seismic waves as they follow a complex path in the propagating medium. To take advantage of the larger sensitivity of scattered waves, Poupinet and Ellsworth (1984) proposed an interferometry method based on coda waves recorded before and after the perturbation. Even if this method allows to retrieve very small changes in the medium, the temporal resolution is generaly limited since the measurements are based the observation of co-located earthquakes (i.e., multiplets).
In Brenguier et al. (2008) and Duputel et al. (2009), we proposed an alternative approach based on the background noise cross-correlations. Cross-correlation of background noise recorded at two seismic stations over a long time-period allows to continuously retrieve the Green’s function between the two observation points. The new approach that we proposed thus allows the continuous monitoring of small velocity changes (< 1%) using the coda retrieved from background noise cross-correlation. This method is particularly adapted to volcanoes since the volcanic erruptions are generally associated with changes of the propagating medium.
Relative velocity changes measured at the Piton de la Fournaise (La Réunion) from January 2006 to June 2007. Variations measured for the station pair bor-fer located in the central part of the volcanic edifice are presented in (A). Grey areas represent periods of eruption, the vertical dotted dark grey line in delimits a major collapse of the main crater (Dolomieu crater) which occured during the April 2007 eruption. Regionalization of temporal changes are shown in (B)-(G). The period taken for regionalisation maps are specified with the arrows and the corresponding letters.
The velocity changes are clearly correlated with the volcano-tectonic activity at the Piton de la Fournaise. The regionalization maps shown in the above figure suggest that the velocity variations mostly affects the central cone of the Piton de la Fournaise which is consistend with data from GPS extensometer and tiltmeter networks. This localization of velocity changes is also suggested by the figure bellow showing the azimuthal distribution of absolute velocity changes 10 days before the July 2006 eruption.
Distribution of absolute relative velocity changes as a function of azimuth for stations pairs containing (A) TCR or (B) fer. The cross-correlation function is averaged over the 10 days before the July 2006 eruption starts.
Dec 2009 New results coming from Mt Ruapehu, New Zealand are now published (see publications page).
Dec 2008 Our paper on the real-time monitoring of velocity changes at the Piton de la Fournaise volcano is now published in JVGR (a reprint is available from my publications page).
Jan 2008 Our paper on the detection of changes in the elastic properties of volcanic edifices using seismic noise cross-correlation is now published in Nature Geoscience (see abstract and reprint in my publications page).
To overcome the limitation of classical f-k analysis (or the equivalent p-f and f-c transforms widely used in near-surface geophysics), we propose an alternative wavefield representation in the group-velocity/phase-velocity (U-c) domain.
Comparison between (left) f-c diagram and (right) U-c diagram computed at f=15Hz (f: frequency, c: phase velocity and U: group velocity). We used synthetic waveforms generated by a surficial source and observed by a linear array. These synthetics were propagated through a simple 1D model of the subsurface. r(k) is the so called array response (here shown at f=15Hz). Black dots indicate the theoretical group and phase velocity values of the fundamental mode (0) and the first overtones (1-2). Crosses indicate the location of the peak maxima for mode 0 and 1. The fundamental mode and the first higher mode are clearly visible on the U-c diagram allowing a measurment of their group and phase velocities for different frequencies. On the other hand, the classical f-c transform is effective in estimating the fundamental mode phase velocity but fails for the first overtone which is too weak to be measured accurately.
Dec 2010 Our paper is now published in Geophysics (doi: 10.1190/1.3341172, a reprint is available from my publications page).
Dec 2009 This work is now accepted to Geophysics and has been presented in AGU 2009 fall meeting. (see publications page).
California Inst. of Tech.