“We should prioritize education as well as training...”

An interview with Mauricio Sacchi

Coordinated by: Satinder Chopra | Photos courtesy: Penny Colton
Mauricio Sacchi

Mauricio D. Sacchi is a well-recognized name in Canada and internationally for his contributions most notably in seismic imaging and inversion. He received a Diploma in Geophysics from The National University of La Plata, Argentina, in 1988 and a PhD in Geophysics from UBC, Canada, in 1996. Since 1997, he has been at the Department of Physics at the University of Alberta, where he has established his industry sponsored consortium SAIG (Signal Analysis and Imaging Group). His works on least-squares migration and prestack data regularization have been particularly interesting and have captured the attention of industry experts worldwide.

Mauricio has developed and taught short courses for the industry and for the CSEG and EAGE societies in the area of seismic signal theory, transform methods for signal enhancement and seismic inversion. With Tad Ulrych, his Ph.D. supervisor and mentor, he wrote the book “Information-based processing and inversion with applications” published by Elsevier.

Mauricio received the prestigious Schlumberger Foundation Research Grant in 2001, and in 2004 his first Ph.D. student, Henning Kuehl, won the SEG’s J. Clarence Karcher Award for his work on least-squares migration. In 2008, Mauricio received the CSEG Meritorious Service Award, recognizing his varied services to the Society and the geophysical community as a whole.

Mauricio is a member of the CGU, CSEG, EAGE, IEEE and SEG societies.

Mauricio was in Calgary a couple of weeks ago and along with Somanath Misra, one of his graduate students, he sat down with us to chat about various things. Following are excerpts from the interview.

Mauricio, I would like you to speak a little about your educational background and your work experience?

I grew up in Brandsen, a town in the province of Buenos Aires in Argentina, where I attended the local public high school. In 1983, I moved to the city of La Plata to attend the public university. I completed a degree in geophysics, “licenciatura” in Spanish, at the Facultad de Ciencias Astronómicas y Geofísicas at the Universidad Nacional de La Plata (Faculty of Astronomy and Geophysics). The “licenciatura” is similar to a combined BSc and MSc; it is a five-year degree plus a thesis. The first three years were occupied mainly with math and physics whereas the last two years were dedicated to geophysics courses. The final degree dissertation was on homomorphic signal processing. Afterwards, I worked as an interpreter for a short period of time for a consulting firm. Interpretation was not my thing (too much geology) and I decided to stay at the University of La Plata. In 1989, a scholarship from the National Research Council of Argentina and a teaching assignment at the University of La Plata let me work on signal processing for 3 years. My supervisor was Dr. Alberto Comínguez. We worked on phase retrieval and wavelet processing using homomorphic operators and blind deconvolution algorithms. During the same time period, Danilo Velis and I started collaborating on seismic bandwidth extrapolation, a topic that is very popular today.

So you worked for 3 years thereafter and then decided to go for a doctorate in Geophysics. What motivated you to go in for this and choose UBC? That would have meant that you would be taking up teaching as a career?

Things were not so good at the university and overall in Argentina. In 1989, there was a period of hyperinflation and a sense that everything was falling apart. I started to look into ways of finding funding to study abroad. At that time universities did not have web pages explaining the process of admission. I had to write letters and wait. On top of that I had no clue about how to find a supervisor interested in my background. Fortunately, I met Sven Treitel during one lecture tour for Amoco Argentina. Sven suggested contacting Tad Ulrych at UBC. Then, the rest happened very quickly. I was admitted to UBC, I married my wife Luciana, and after six months we moved to Vancouver. My son, Federico, was born in Vancouver a year after commencing my PhD. My dissertation was finalized in 1995 and my daughter, Bianca, was born the next year. The children helped us to feel at home in Canada and this is why we decided not to return to La Plata. After a short postdoctoral position at UBC, I accepted joining the Department of Physics at the U of A where I have been a faculty since 1997. A teaching position had never been in my mind. A research career in industry was my goal. I applied for industry jobs and for my current position at the U of A. Industry said no, university said yes; a faculty position at one of the top universities in Canada was something not planned, but it happened. I am a very lucky man. I had never seen myself as a teacher, my pedagogical aptitude was subpar to say the least; it happened that teaching was one of my duties, which I incidentally found enjoyable especially when it comes in small amounts.

Since you came to the U of A, you have essentially stayed there. Could you tell us how you decided on that?

The Department of Physics at the U of A is a civilized place. My colleagues never made me feel a second-class citizen for working on non-fundamental physics. They were (and are) very collegial. I had the enormous opportunity of supervising top-notch students, the type of students that are only found in really good schools. Graduate students are the backbone of any serious research program and our department provides an excellent environment to attract students interested in my field. The geophysics group at U of A has a good balance of applied and non-applied science that, in my opinion, is essential to delivering a good graduate program.

Support from the University in the form of start-ups funds, and the freedom to work independently and to recruit my own graduate students were key to transition into a permanent tenure position. In addition, my teaching duties were confined to upper division geophysics courses in applied and global seismology, time series analysis and inversion, and my administrative duties were minimal. Friends and colleagues in Calgary were also instrumental to the development of my research program. Mike Perz at Geo- X, Bill Goodway, Lee Hunt, Dave Copper and Dave McKidd at the old PanCanadian, and Scott Cheadle and Helmut Jakubowicz at Veritas provided initial funding to support graduate students. My chances to start a research program in applied seismology would have been slim without initial help from industry. More companies joined the group later and we now count 10 sponsors. In short, to answer your question: support from the U of A, support from colleagues in industry and access to good graduate students have made it possible to run an active research group and, this is the main reason I have stayed at the U of A.

Who have been some of your mentors? Tell us about your association with Tad Ulrych. He was your supervisor for Ph.D., but you have had a productive association with him, publishing a book with him besides many research papers?

I met many extraordinary people during the initial years of my career. Dr Alberto Comínguez at the National University of La Plata taught me the fundamentals of signal analysis. Thanks to him I arrived at UBC well prepared to work on seismic signal processing. Tad Ulrych, my supervisor at UBC and friend for life, changed my world. Conferences in North America and Europe were not a problem long as I had something to show! He also started an interesting association with industry via CDSST (Consortium for the Development of Specialized Seismic Techniques) that produced vigorous research in transform methods, statistical signal processing and time-frequency methods. Tad has constantly stressed that one needs to work on problems that are interesting and not necessarily mainstream. CDSST was all about ideas; industrial partners were in charge of finding how to use these ideas. CDSST with Tad and fellow graduates Xin-Gong Li, Ken Matson and Danilo Velis was an interesting intellectual journey and a point of reference for my current group.

Besides my mentors, I am also indebted to my wife who gave up a law career in Argentina to follow me to Vancouver and then to Edmonton. My father also helped us in many ways. I would not have been able to survive graduate school without their support.

What personal qualities do you think enabled you to achieve the professional status that you enjoy today? Is it hard work, ambition, or anything else? Do you also think you had a firm grinding in mathematical concepts that has kept you in good stead or is it something else?

Collaboration with Tad Ulrych, Daniel Trad and students at the U of A led to interesting ideas that were followed and implemented by industry. I don’t have a formal education in mathematics; I do have some mathematical intuition and a good feeling at the time of developing algorithms and that might have helped with some of our projects. It is also true that I have the privilege of working with students who have completed exceptional PhD dissertations: Henning Kuehl, Bin Liu, Juefu Wang, Somanath Misra (with us today), and, more recently, Mostafa Naghizadeh.

Dr. Mauricio Sacchi, Satinder Chopra, and Somanath Misra
Interview gathering (Right to Left): Dr. Mauricio Sacchi, Satinder Chopra, and U of A PhD graduate Somanath Misra, now at Arcis.

Tell me about Mauricio, the person- i.e. your habits, your likes/dislikes, etc.?

My son strongly insists that I am the “Dr. House” of the Physics Department. People also tell me that I am very contradictory. I was told that this is because I am a Pisces. No, seriously, I like working with graduate students and to see them evolving into mature scientists. I think that the latter is what keeps me going at the University. As for dislikes: people who take themselves too seriously.

I notice from your CV that you are also working with your students on applications of geophysical methods in other areas like acoustic imaging of bone density and non-exploration seismology. Tell us about this and other such applications?

Professor Lawrence Le (Radiology and Physics) and I have been collaborating on problems associated to bone imaging and characterization. We co-supervise two graduate students. My role is to help with the signal processing and imaging side of the research. Lawrence’s laboratory is currently involved in building sensors and methodologies to characterize bones and to aid osteoporosis diagnosis.

With our global seismologist, Professor Jeff Gu, we are working on the application of signal processing methods to image upper mantle discontinuities. My role here, again, is to provide assistance to adapt exploration seismology methods to problems in global seismology. With Jeff and former graduate student, Yulin An, we have successfully applied high-resolution Radon Transforms to isolate upper mantle phases. In summary, whenever possible, I try to find ways to collaborate with people interested in imaging and signals at all scales and applications.

What has been the most memorable moment in your professional life? Also, tell us about some of the successful landmarks in your geophysical career? This could be your most important contribution to geophysics, which you are really proud of?

During my PhD, I managed to reconstruct highly decimated data using a sparse inversion algorithm that I wrote in F77. This was based on ideas discussed with Tad Ulrych and his former collaborator Collin Walker. The idea was to use sparse priors to obtain high-resolution estimates of the power spectral density of a time series. We turned the problem into a reconstruction algorithm and used it to recover highly decimated signals. Tad and I become interested in sparsity or parsimony (as Tad calls it) and apply sparse inversion to several problems associated with seismic data processing. The idea is quite simple and we elaborated it in very intuitive terms. Seismic data can be represented in a new domain (the transformed data) that in ideal conditions should be sparse. Incomplete data, on the other hand, will transform into a less sparse domain. Algorithms to retrieve a sparse version of the transformed data were used to estimate the data that one would have recorded under ideal conditions. The “sparsity trick” was applied to the inversion of Radon and Fourier operators. It is important to stress that our work on Radon Transforms was inspired by contributions by Dan Hampson, Jeffrey Thorson and Jon Claerbout in the mid eighties. There was a decade of little activity on the subject until the publication of our (Sacchi and Ulrych) Geophysics paper in 1995. In some way, our 1995 paper refloated interest on the topic. The frequency domain High Resolution Radon Transform immediately caught industry attention and important progress was made by people working in seismic data processing. For instance, Philippe Herrmann and co-workers, at CGG, added anti-alias capability to the high-resolution Radon Transform.

Similar ideas were also used to regularize multidimensional surveys. In particular, Liu Bin in collaboration with Daniel Trad, implemented the first (to my knowledge) interpolation/reconstruction scheme for 5D data. Preliminary findings were shown in Bin Liu’s thesis. Daniel played an important role in turning Bin Liu’s academic research into an industrial application. Because I was not trained as a mathematician, I never thought about the conditions that are required to obtain the correct data reconstruction. Today, thanks to the work of many clever mathematicians, sparsity-promoting algorithms (as they are often called) are part of a hot research field named Compressive Sensing. The field pays particular attention to optimality conditions for data recovery. Felix Herrmann’s group at UBC is very active in this area and has provided very interesting applications to data processing, modeling and inversion.

What are your aspirations for the future?

As per career aspirations, I would like to continue to run our small consortium, a nice way to fund graduate students and to transfer ideas to industry. I would definitely like to move into some sort of a shared academic and industrial position that would allow me to pay more attention to the research carried out by our group. It will require a new way of dealing with industry/university relationships and I am very open to all ideas! To same extent, I like the intellectual freedom provided by the university system, but I realize that most important advances in applied seismology do happen in industry.

You have a position of a professor of geophysics at University of Alberta. Please tell us about, the type of work that is being done in your group and how many people are engaged in doing that.

I am currently supervising 10 graduate students and one postdoctoral fellow. Our funding comes from industry under the umbrella of SAIG (Signal Analysis and Imaging Group). I have students working on sampling, regularization/interpolation, imaging with regularized least-squares migration, coherent noise attenuation, Bayesian AVO inversion and microseismicity.

For instance, Ismael Vera Rodriguez is building a system for automatic detection of microseismic events using ideas from the field of Compressive Sensing. Mostafa Naghizadeh has continued the regularization/interpolation line of work initiated by Liu Bin and developed robust f-x interpolation algorithms for irregular data. Mostafa’s recent PhD thesis shows field examples of 5D (time and 4 spatial coordinates) data regularization using prediction filters. It is a very fine extension of the f-x interpolation method of Simon Spitz to the irregular and multidimensional sampling case. Sam Kaplan is working on wave equation regularized migration and its regularization and preconditioning aspects. He is also looking into possible extensions to simultaneous source imaging. Wubshet Alemie has continued the steps initiated by my good friend and former PhD student Somanath Misra who is here today. Wubshet is working on Bayesian AVO analysis and has developed an algorithm that allows us to retrieve highresolution estimates of correlated AVO attributes. Amsalu Anagaw is about to complete a MSc. dissertation on studies of waveform inversion via total-variation regularization and adjoint-state methods. Vicente Oropeza is working on eigen-decomposition methods for coherent noise attenuation.

You and your students have won many awards. Tell us about some of them. Though all awards are recognition of your contributions, which are more close to your heart?

They are all close to my heart and I must say again that I was lucky in having the opportunity of supervising great graduate students who have all, in one way or another, made important contributions to seismic data processing.

I notice that you are on the editorial board of several journals and committees of professional societies. Tell us why do you do all this?

I try to allocate time during the week for different activities, sometimes it does not work and I end up working late at home. Being part of the editorial board of Geophysics is interesting because one can see how the peer review system works from the inside.

Some of the areas that you have worked in are seismic imaging (least-squares wave-equation migration); transform methods, seismic data reconstruction using interpolation, wavelet estimation. Could you tell us briefly about the challenges we face in each one of them and how you have tried to address some of the issues?

Least-squares regularized migration is a nice way of imposing a priori constraints to an image while finding the image that satisfies the data. Efficiency is the main issue. The cost of least-squares migration is the cost of two migrations per iteration and a good solver often needs 4-10 iterations to find an optimal solution. The real benefit lies in its capability of constructing images with reduced sampling and migration artifacts. The latter was the main research focus of former SAIG graduate students Henning Kuehl and Juefu Wang. Juefu has continued to work on this direction at Divestco. He has devised a preconditioning strategy that preserves structural dip and, at the same time, attempts to cancel artifacts. Juefu will present nice field data results at this year’s SEG meeting.

In the area of seismic data reconstruction we have many challenges. I do believe that these are good times to work on data reconstruction problems. Companies are engineering clever algorithms for multidimensional regularization. The key point is to work with multi-dimensional algorithms where the reconstruction of undersampled data in a given direction is assisted by well-sampled data in other directions. This year’s EAGE Regularization Workshop in Amsterdam was a good showcase of industrial applications of multidimensional regularization. So far, it seems that regularization and interpolation have been mainly guided by processing requirements (Migration, AVO and AVAz studies, SRME, etc). This might be the right time to look into interpolation/regularization as a way to also decrease acquisition costs.

Satinder, you have also mentioned, my hobby, wavelet estimation. This is a longstanding problem, a problem with one equation and two unknowns. Therefore, wavelet estimation algorithms either make a phase assumption or impose stringent statistical conditions on the reflectivity series. On top of that, we have the issue of the validity of the convolution model. It is an interesting problem and, no doubt, a great problem to learn the basis of statistical signal analysis. It can be discouraging, however, to find out that our best intentions might end up with algorithms that only work with examples that partially mimic real scenarios.

What other type of problems are you working on?

A current challenge is to build from scratch a comprehensive noise attenuation toolbox that includes all the tricks I learned during the last 20 years. In addition, a project of chief interest is to develop methods to attenuate coherent noise in conjunction with survey regularization. Our regularization/interpolation algorithms assume data free of large amplitude coherent noise. Noise removal, on the other hand, requires well-sampled data. It seems to me that acquisition/sampling and processing strategies need to be studied in concurrence.

I have also been working on adaptive filtering to cope with spatially varying dip in t-x and f-x noise attenuation methods and adaptive subtraction.

I recently developed an algorithm that represents seismic data using a sparse Gabor representation that one can use for trace regularization. With Sam Kaplan and Tad Ulrych we have used independent component analysis (ICA) to construct basis functions similar to local Radon bases. The goal is to generate a library of local waveforms to represent seismic data. A similar project has just been completed by Kenny Kocon, an undergraduate summer student funded by NSERC, and we are now looking into applications for random and coherent noise removal. I am also becoming more and more interested in Compressive Sensing and experimental design as a way to find fast solutions to inverse problems.

You conduct a short course on ‘Statistical and transform methods for seismic signal processing’, for the oil and gas industry. What all do you teach in this course and how is it useful and to whom?

This is a course that starts with basic signal processing concepts, convolution and filtering. The least squares inverse filtering is used as intro to inverse problems and regularization (pre-whitening) etc. The course continues to build on signal processing and inversion concepts of more complexity and culminates with large-scale inverse problems like the ones encountered in least-squares regularized migration and multidimensional seismic data regularization. The course is useful for people interested in the development of algorithms for seismic data processing. It might also help to understand some of the technologies that are currently used to process seismic data. The course has evolved into a new one called “Processing, inversion and reconstruction of seismic data” that emphasizes recent work on reconstruction and regularization of seismic surveys.

You have written a book on statistical and transform methods. What drew you to book writing?

Pieces of this book are part of notes I developed to teach a course on geophysical signal processing at the U of A and notes for short courses for industry. Some of the chapters of “Statistical and Transform Methods” evolved into a book called “Information-Based Signal Processing”. The latter is a book that I published in collaboration with Tad Ulrych that not only covers transform methods but also information theory, Bayesian inference, eigen-analysis of seismic data etc. Our intention was to provide a survey of interesting (not often discussed) problems in signal analysis and inversion and their application to exploration seismology.

I would like to know your perception about writing; is it sharing of information, experience, knowledge or something else?

Writing is a simple way of remembering what I want to say when I am teaching. All my notes are written with a word processor that can easily insert equations by typing (Latex). It gives me some sense of security when I teach. Yes, I do think that in my case it all starts as a way of sharing information with my students. I definitely dislike using power point presentations for my lectures. It forces one to provide initial and final results and to omit all that goes in the middle and, important details that are usually in the middle of the discussion.

You were at Delft on sabbatical visit in 2003. Tell us about your experiences there.

It was an extraordinary experience, the city of Delft is a nice place to live. The sabbatical offered time to read and catch up with several topics. My hosts were Dr. Eric Verschuur and Dr. Dries Gisolf at the Delphi Consortium. I primarily worked with Paul Zwartjes (a former PhD student of Dr Gisolf) on problems regarding seismic data reconstruction. Paul devised interesting interpolation schemes based on sparse priors that complemented quite well with our research at the U of A. A sabbatical year was also an opportunity to travel to companies in France, Italy, Norway and England to teach short courses and meet fantastic people with whom I have developed a long-lasting friendship.

What are your other interests?

Reading is my hobby. I have just finished reading Patrick French’s biography of VS Naipaul and I am currently reading an interesting book about post-independence wars in Argentina.

What would be your message to young undergraduates who have decided geophysics as their future profession?

Our young bright undergraduates are often too concerned with the profession and with getting the right training for the profession. My advice is to use your years at the university to learn the basis of our discipline (math, physics, geology) and don’t be too concerned about the profession; you will only narrow your learning experience and make yourself less flexible for those times when the profession goes through bad times. We should start to prioritize education over training. This is not an easy task; many of our geophysics programs are engineered to train professionals. You will have plenty of time, when joining the workforce, to learn what the profession is all about and to be trained to become a professional. A good education should also have abundant room for sciences that are peripheral to the profession and, for the arts and the humanities. It is difficult to predict the type of geophysicists that our society will require in two decades. Therefore, students should be prepared for a future that may or may not have today’s opportunities.

One last question, do you think I missed out on any question, that you expected me to ask and I did not?

I was afraid you were going to ask if I like the song “Don’t Cry For Me Argentina” (which I don’t like, by the way). No, seriously, you did not miss any question. Thanks for the invitation to the interview.


Share This Interview