Laurent Sirgue
 | |
| BSc | Physics |
|---|---|
| MSc | Geophysics |
| PhD | Geophysics |
| BSc university | University of Versailles |
| MSc university | Strasbourg University |
| PhD university | University of Paris-sud XI |
Laurent Sirgue joined Total France in 2009 as a geophysicist. He has been conducting research on the topic of full waveform inversion (FWI) for nearly 15 years. Laurent received a BSc in physics from Versailles University in 1995 and an MSc in geophysics from the University of Strasbourg in 1997. He joined CGG in 1998 as a research geophysicist. In 1999, CGG sponsored Laurent's 16-month National Service, which took him to Queen's University, Kingston, Canada, where he worked with Gerhard Pratt on FWI. In 2003, Laurent received his PhD from the University of Paris-sud XI. His dissertation research also focused on FWI. CGG sponsored Laurent's doctoral research in collaboration with Professor Pratt. As a postdoctoral researcher at Total (Pau, France), Laurent conducted research on reservoir characterization. In 2004, he joined the depth-imaging group at BP America where his work focused mainly on FWI. Laurent has received a number of awards for his research including Honorable Mention for Best Paper in Geophysics in 2004 and the 2010 Bonarelli Award for best oral presentation at the EAGE Annual Conference.
Contents
SEG Virgil Kauffman Gold Medal Award 2016 [1]
R. Gerhard Pratt and Laurent Sirgue are recognized for their development of full-waveform inversion. Evidence provided by notable letter writers provides convincing proof that Pratt and Sirgue are deserving of this award. The letter writers want to credit the people who persistently worked on the idea of full-waveform inversion and eventually made the concept succeed using real data.
Biography Citation for the SEG Virgil Kauffman Gold Medal Award 2016
By John T. Etgen and Andrew Brenders
The long journey of bringing full-waveform inversion (FWI) to industrial application is a story of persistence, perseverance, and timing. Perhaps it was fate, or maybe just coincidence, that brought these two gentlemen scientists together at the right juncture in our quest for better velocity models both for imaging and as standalone descriptions of the subsurface. Whatever the case, their individual efforts — made stronger and more complete by their work together — made the true and definitive difference that brought FWI out of the academic world and into industry. The course of our field has changed in a profound way, with industry finally accepting that solving the single-most difficult issue in seismic imaging is in fact possible.
Of course, posing geophysical parameter estimation as an inverse problem was not a new idea when Gerhard started his career, first as a grad student and then as a young lecturer and professor at Imperial College. Since the ranks of the “inverters” was a small community, Gerhard had the privilege of working with many of those who made seminal contributions to geophysical inverse problems, the likes of which included Tarantola, Lailly, Mora, Shin, Marfurt, Woodward, and others. It was Gerhard’s work on real ultrasound data, implementing what was essentially a form of diffraction tomography in the frequency domain, that took him on a path different from the direction many others in the inversion community were following.
Those early successes on real data, and the commitment to working on field data instilled in him by his thesis advisor Michael Worthington, helped shape his understanding of how to make a genuinely practical FWI, before the term FWI was even coined. Gerhard deviated from some of the academic crowd of the day by embracing the acoustic approximation, when at the time it was the general consensus that one had to reproduce the exact waveform in all detail. He is a firm proponent of the “80% solution,” and continually challenges his students to move on from synthetic tests and apply the method to field data. This philosophy was instrumental in later building the bridge to industry. We recall many times that Gerhard would shy away from the term “full-waveform inversion” and preferred to use the term “waveform tomography” in his presentations to better explain how to make the method succeed based on the data it was given. I (John) always found a way to attend his SEG presentation on this topic, even in those days when inversion talks only drew small audiences and there was at most one session at the Annual Meeting.
As the 1990s progressed, Gerhard and his students and colleagues created a growing body of advances in frequency-domain FWI. Even though Gerhard had industry sponsorship, in the 1990s few in the industry were giving FWI much chance of success or usefulness. Gerhard recalls instances where even showing a title slide with the words “waveform inversion” would elicit groans from industry audiences, as most thought the method too expensive and impractical. In 1998, Gerhard returned to Canada to take a position at Queen’s University, where he recruited and nurtured a new and talented group of students, notably including Laurent Sirgue.
Laurent’s first major contribution to the science of FWI was to demonstrate and justify why a limited set of frequencies can, in fact, recover the wavenumber spectrum of the velocity model. The time-domain FWI community had always held the frequency-domain FWI community in disdain, partly because they did not have rigorous justification as to why a sparse set of frequencies would work. To overcome these objections, Laurent and Gerhard published a paper in Geophysics in 2004 showing how to choose frequencies for FWI and received Honorable Mention for best paper in Geophysics. This paper firmly established FWI as a relevant technique in academic geophysics, although even into the early 2000s, real data tests were few and far between, and none existed at industrial scale.
The watershed event occurred at the 2004 EAGE in Paris, at an industry sponsored workshop to “capture a snapshot of the state of the art in velocity estimation,” featuring a challenging 2D acoustic synthetic data set. The data alone were distributed to any and all comers; the model was not. Many research and processing groups from industry and academia participated. One seismic processing company confided in me (John) that they had a team of four people working full time for a month, using a vast amount of computational resources to craft their company’s entry, simply by guessing and testing. While it was true that some of the large industry players created good images, it was clear that they exhausted their capability and still hadn’t recovered all of the model’s complex aspects. All eyes were drawn to a result created by a small research group from Queen’s University. With nothing more than a frequency-domain FWI code running on a single CPU and the grad student efforts of one of us (Andrew), FWI created a model that at least recognized (if not solved) all of the “dirty tricks” embedded in the model. This result was so instrumental to our seismic-imaging group that we immediately wanted to hire students from Gerhard’s program.
So a little less than a year later, Laurent began working for the industry and showed how FWI can solve velocity-model estimation problems using just a few frequencies, and given low enough frequencies, a decent starting model, and long enough offsets. Crucially, Laurent had excellent intuition about what combinations of those quantities would more or less guarantee success. The industrial task was, of course, nontrivial. First, real problems were 3D, placing the computations out of reach, and velocity model determination projects were tough, involving poorly imaged salt or shallow gas zones.
The next big advance came when Laurent discovered a way to use time-domain propagators in frequency-domain FWI. Uwe Albertin showed Laurent how efficient time-domain propagators were and lamented that we had to solve the Helmholtz equation, the prime limitation of applying frequency-domain methods to 3D. Laurent examined some modeling codes and realized that the time-stepping loop was a very natural place to apply the discrete Fourier transform on the fly.
Almost immediately, we had a working, easily scalable, 3D frequency-domain FWI code. We used it in secret for a while to study the salt problem, which continued to look difficult. But, at my (John’s) suggestion, we tried it on resolving low velocities in gas-invaded zones above Valhall field. The result was absolutely beautiful and gave stunning details of the gas zones. When that model was used to image the ocean-bottom data set, the migrated image gave unprecedented clarity where there was previously no image. The result was kept secret for a while, but because of the implications on monitoring shallow hazards, the result (without saying what it was exactly) had to be shown to the other Valhall partners. Of course, some of them recognized what it was, and the race to apply FWI to industrial problems was now on. The Valhall results were eventually published, as well as the results on many other imaging-challenged major oil and gas fields. We wonder if Gerhard and Laurent ever would have predicted that FWI would catch and surpass migration as the largest category in the technical program at the SEG Annual Meeting.
2014 SEG Honorary Lecturer, Europe
Full waveform inversion of seismic data: Investigating the Earth for high-resolution velocities and more...
The principles of full waveform inversion (FWI) have been established for about 30 years. This approach aims at finding an Earth model that explains the seismic data. This is achieved by estimating the physical properties of the subsurface (Vp, Vs, density, attenuation, anisotropic parameters, etc.) within an inverse problem that aims at minimizing the difference between seismic data acquired in the field and predicted data modeled in a computer. The ultimate goal of FWI is therefore very ambitious as this multiparameter inversion is a highly nonlinear and ill-posed problem. In addition, it requires accurate and efficient modeling of the wave equation which makes it challenging to apply to real, industrial-scale, 3D problems.
In recent years, rapid improvement of supercomputers and breakthroughs in numerical computations of various forms of the wave equation allowed the academic community and the seismic industry to apply this technique in 3D.
Initial applications of 3D FWI made the assumption of acoustic-wave propagation, thus ignoring viscoelastic effects and mainly focusing on the recovery of a single parameter: acoustic velocity (Vp).
While the recovery of this unique Earth model parameter seems to fall short of the stated ambition of FWI, it is at the heart of prestack depth imaging commonly used by the seismic industry. In order to obtain an image in depth, a velocity model must be estimated prior to performing prestack depth migration (PSDM) which produces the final image for geologic interpretation.
The process of velocity model building is a critical stage of this workflow and is conventionally performed using ray-based reflection-tomography techniques. Velocity models derived by such methods are typically low-resolution and are used only for PSDM.
In this presentation, I will first review the basic theory of FWI. I will then define, by means of illustrative synthetic examples, the key parameters that play a role in the success of FWI (data and model requirements). This will lead to the understanding of why FWI has enjoyed its most success in the recovery of shallow velocity anomalies.
I will continue by showing real data examples that demonstrate the potential of FWI to generate high-resolution velocity models which may improve the images produced by PSDM. Because of the highly detailed nature of the velocity field produced from FWI, it also contains valuable information that can be used directly for geologic interpretation. I shall finally discuss ongoing research and the evolution of FWI across the academic and industrial communities.
This lecture is intended for a large audience and no background knowledge about FWI is needed.
Honorable Mention (Geophysics) 2004
Laurent Sirgue and R. Gerhard Pratt received 2004 Honorable Mention (Geophysics) for their paper Efficient waveform inversion and imaging: A strategy for selecting temporal frequencies[2]
References
External links
| find literature about Laurent Sirgue |
|
|
|
|
|
|
|
