Sergey Fomel strives to improve the processing and analysis of seismic images. He also created the Madagascar software package. Photo: Sasha Haagensen.
Sergey Fomel Advances Seismic Data Analysis
By Marc Airhart Nov. 07, 2007
Sergey Fomel grew up in Akademgorodok, literally “Academic Town,” a scientific hub in Russia’s third largest city, Novosibirsk. Surrounded by a forest of birch and pine trees, tens of thousands of scientists live and teach at Novosibirsk State University, and do research in any of 37 research institutes. The town, which was founded in the 1950s by the Soviet Academy of Sciences, is its own little universe, complete with homes, stores, hospitals, restaurants, clubs, and cinemas.
Fomel’s parents were physicists. Most of their friends were scientists. His classmates were children of scientists. Most went on to study science at the university.
“There was never a question that I would go into science,” said Fomel. “It was just a choice between disciplines.”
Still, as a child, Fomel wasn’t a junior scientist. He didn’t have a rock collection or perform chemistry experiments in his bedroom. He preferred cross-country skiing. But he was good at math.
In Russia, students must choose a career path when they graduate from high school.
Fomel was 16 years old when Sergey Goldin, a friend of his father and a brilliant and charismatic professor at the university, helped nudge him towards geophysics.
“He used to say that geophysics is physics of the real world,” said Fomel. “If you study physics, you can focus on abstract things like elementary particles. But geophysics is about the real Earth. It’s harder because you can’t replicate it in the lab, so that was an attraction to me.”
Akademgorodok, a scientific hub founded in the 1950s within Russia’s third largest city, literally means “Academic Town.”
In 2002, Fomel came to work as a research associate at The University of Texas at Austin’s Bureau of Economic Geology.
“My specialty is exploration geophysics, the application of geophysical methods to petroleum exploration,” said Fomel. “And UT is a very good environment for that. So I came to the Bureau, which is an organization with many excellent connections to industry.” He also appreciates the flexible research groups at the Bureau that allow interactions between people with different backgrounds and interests.
While at the Bureau, Fomel created Madagascar, an open-source software package used primarily for processing and analyzing seismic data. The package is part of a new movement in computer-aided science, which seeks to make computational experiments more transparent and allow independent researchers to reproduce and confirm other’s results.
In September 2006, the Society of Exploration Geophysicists rated three of Fomel’s presentations among the top 30 given at its annual meeting. There are typically 600 or 700 presentations given each year. One of the presentations, titled “Local Seismic Attributes,” demonstrated a new way to extract additional information from seismic images. In particular, it showed a way to reconcile two different kinds of images—those produced with shear waves and compression waves—for a better characterization of the Earth’s interior. This paper won the best poster presentation award.
Now, in the fall of 2007, Fomel has begun a dual appointment at the Department of Geological Sciences, where he will spend a third of his time teaching courses and mentoring students as an associate professor. The other two-thirds of the time, he will continue his work at the Bureau.
From Theory to Application
Fomel acknowledges that three mentors have been especially important in his career.
This seismic image from the North Sea was created using a new technique proposed by Fomel and colleagues at U.C. Berkeley for converting from time to depth.
He was greatly influenced by Sergey Goldin, the man who first set him on the geophysics path. After graduating from Novosibirsk State University with a geophysics degree in 1990, he went to work on theoretical problems with Goldin at the nearby Institute of Geophysics.
“His knowledge of geophysics was very broad and deep,” said Fomel. “You could go to him with any question and he would know the answer. He was so far ahead of his students and that was intimidating.” He developed a reputation as a top scientist in his field. Special conference sessions have been organized in his honor.
Goldin, who passed away last year, taught himself how to paint, draw, write poetry, and play a stringed instrument called a domra, as well as write and direct theatrical productions.
“The way he explained it was that he felt the need to exercise a different side of the brain,” said Fomel. “He thought that one can not only do science all the time.”
Fomel admired and respected his advisor, but actually grew the most as a scientist when Goldin was away at Stanford for four months and later in Brazil for a year.
“Goldin’s biggest gift to me was actually going away,” said Fomel. “I went through a lot of frustration but finally learned how to work independently.” Fomel became interested in going beyond theoretical work.
“I wanted to break away from theory and work on practical applications and that’s mostly why I went to study at Stanford and changed my specialty,” said Fomel.
At Stanford, he studied under Jon Claerbout and earned a Ph.D. in geophysics in 2001.
“Jon is a legendary person in our profession and the scientific father of about 50 graduate students,” said Fomel. Some went on to become professors at places such as MIT, Caltech, Stanford, and Colorado School of Mines. Most went into the petroleum industry and became industry leaders.
“Unlike Goldin, you could go to Claerbout with a question and he would not give you an answer even if he knew it,” said Fomel. “That’s partially why many of his students became so successful and independent.”
It was at Stanford that Fomel developed an interest in reproducible computation. Claerbout and his students at the Stanford Exploration Project—an industry-funded academic consortium designed to improve the theory and practice of constructing 3-D and 4-D seismic images of the Earth—pioneered the application of reproducibility to computational geophysics. This concept is at the heart of Fomel’s Madagascar software package, which was publicly released last year.
Why Madagascar?
The mind strains to find a connection between Madagascar, a large island nation off the southeast coast of Africa with many plants and animals found nowhere else on Earth, and a piece of geophysical software used to analyze seismic data. And with good reason.
The software had originally been dubbed Regularly Sampled Format, or RSF. Only no one could remember the name or the abbreviation. They would remember it as RFS or some other almost-correct abbreviation. So Fomel decided to find a more memorable name.
“I decided to look at airport codes, those three letter codes that represent airports around the world,” said Fomel. “Maybe there was one with RSF.”
As it turns out, there isn’t. Undaunted, he noticed there was an RSE (in Australia) and an RSG (in Brazil). If there were a mathematics of the alphabet, RSF would be the average of these two places.
“So I drew a straight line between the two and the middle happens to be in Madagascar,” said Fomel. And the name stuck. “At least people can remember it,” he added.
Do That Again
Scientists tend to be a pretty skeptical bunch. When they read published results from an experiment, they want to be able to repeat the experiment and confirm the outcomes. Consequently, they publish their results in such a way that everything can be reproduced, describing exactly what they did and how they did it.
Computers have become an indispensable tool in science. They allow scientists to carry out entire experiments in silico, without labeling a single test tube or releasing a single sensor into the ocean. That ability is leading to amazing breakthroughs from medicine to astrophysics to geoscience. It also has scientists struggling to verify each other’s work.
Fomel, who grew up in Novosibirsk, Russia, earned a geophysics degree from Novosibirsk State University.Image: NSU main building.
In this new landscape, a reviewer has to have the same software—not just the same software, but the same version—as well as the same initial data, and they must use all the same settings and methodology of the original scientists. And yet the traditional medium for publishing results—a scientific paper just a few pages long—can’t possibly give a researcher everything they need to replicate this new class of experiments.
“As it is now often practiced, one can make a good case that computing is the last refuge of the scientific scoundrel,” said Randy LeVeque, a prominent mathematician in a keynote presentation at an international mathematics conference in 2006. He complained that published results of computational experiments often lack detailed descriptions of the methods used. “Scientific and mathematical journals are filled with pretty pictures these days of computational experiments that the reader has no hope of repeating,” he said.
Fomel designed Madagascar as an antidote.
“Reproducibility means maintenance,” said Fomel. It means maintaining computational experiments together with data sets and past versions of the software. It means documenting how the software changes and repeating experiments on newer versions to see how they compare with earlier runs. Fortunately, the rise of the Internet has made sharing data and software much easier.
By using an open-source model, researchers can look under the hood and see exactly how the software works. Developers around the world collaborate to improve Madagascar. A similar approach has led to many well-known software products such as the Linux operating system and the Firefox Web browser. In addition to Fomel, contributors in this ongoing development include Jim Jennings, formerly of the Bureau, now at Shell; Paul Sava, formerly of the Bureau, now at Colorado School of Mines; Joe Dellinger at BP; Ioan Vlad at Statoil; Gilles Hennenfent, Henryk Modzelewski, and Colin Russell at the University of British Columbia; and others.
Computational reproducibility is now appearing in other fields such as statistics, bioinformatics, and econometrics. For Fomel, though, the focus remains seismic imaging.
Late Arrivals
After graduating from Stanford, Fomel took a postdoctoral position at Lawrence Berkeley National Laboratory and worked with Jamie Sethian, the third scientist to have an influence in his early career.
“Jamie is also a legend,” said Fomel. “He is a true applied mathematician who loves difficult applications such as seismic imaging.”
Sethian, head of the mathematics department at Lawrence Berkeley, earned the Norbert Wiener Prize in Applied Mathematics “for his seminal work on the computerrepresentation of the motion of curves, surfaces, interfaces, and wave fronts, and for his brilliant applications of mathematical and computational ideas to problems in science and engineering.”
“My postdoc appointment at Berkeley was one of the happiest times in my professional career, because with Jamie’s help, I could go outside of geophysics and from a bird’s eye view try to determine the most important geophysical problems to work on,” said Fomel. “Many of the ideas that I develop today came out of that experience.”
When geophysicists do seismic imaging, they often encounter a problem in which the returning sound waves travel multiple paths. Consequently, different parts of the wave arrive at different times. An example of a medium in which this might happen would be a soft material surrounded by a harder material. Because sound travels more slowly in the soft material, some of the sound would bypass it and travel faster through the harder rocks.
“It would be easy if you could ignore all late arrivals, but that doesn’t produce the best image,” said Fomel. He and Sethian developed a technique that computes the second arrival and yields a clearer image. Their paper appeared in the Proceedings of the National Academy of Sciences. With support from the petroleum industry, this work continues now at the Bureau.
Professor Fomel
Fomel has taught courses in seismic migration theory at Novosibirsk University, numerical analysis at the University of California at Berkeley, and seismic wavefield imaging at UT Austin. So he’s quite comfortable in the role of professor. This fall, in his dual appointment with the Department, he is again teaching a course on seismic imaging.
“I’m excited,” said Fomel. “I really enjoy teaching. The previous times when I did teach, it wasn’t really part of my job duties, but to me it was a joy. Sometimes when you try to explain something to students, you find holes in the existing research or you find that you really don’t understand something. So it can actually benefit your research.”
He has reorganized his seismic imaging course to be more accessible.
“I discovered that when you try to explain it to people outside of our field, you start to run into some inconsistencies and logical gaps,” said Fomel. “So I’ve tried to determine how to tell a consistent story without too much jargon and without resorting to saying something is true because an authority says so. That’s the challenge.”
For more information about the Jackson School contact J.B. Bird at jbird@jsg.utexas.edu, 512-232-9623.
The software had originally been dubbed Regularly Sampled Format, or RSF. Only no one could remember the name or the abbreviation. They would remember it as RFS or some other almost-correct abbreviation. So Fomel decided to find a more memorable name.