Tributes to Bruce Bolt

David Brillinger

Presentation at Memorial Service for Bruce A. Bolt

29 July 2005, Faculty Club, UC Berkeley

My name is David Brillinger.

Bruce Bolt was a dear friend, colleague and collaborator, and it is a great honor to be invited to speak in his memory today. There is so much to say. I start with this room. It is associated with many wonderful events in the Bolt family’s and their friends lives – weddings, birthdays, banquets, lunches, talks and more.

Bruce and I first met in 1968 or 1970, either when I had a visiting appointment from London or had been appointed in the Statistics Department. I had been doing research on distinguishing nuclear explosions from earthquakes and Hal Thirlaway of the UKAWE suggested that I look up Bruce. I did and Bruce welcomed me to Berkeley.

We found we had much background and many interests in common. Perhaps the most amusing is that we both knew Toronto well. I, the Canadian Toronto, and Bruce, the Australian Toronto. In case you are wondering about the Australian Toronto it was named for a famous Canadian rower of the late 1800s, Ned Hanlan. Bruce had learned his beloved sailing on Toronto’s Lake Maquarrie.

I want to talk about the basis of the scientific philosophy driving Bruce’s research and one of the collaborative experiencee that I shared with him.

Bruce’s conversations and writings were filled with references to the renowned scientists Keith Bullen and Harold Jeffreys. These men were both mathematicians and seismologists, and for each their research was driven by the scientific method and the tools of applied mathematics. They were Bruce’s mentors and they launched him on a magnificent career. Bruce implemented their programs and surpassed their work in important aspects including public service, education, and breadth of application.

Bullen was concerned with applied mathematics and the scientific method generally. Bruce quoted him as writing, “The distinctive feature of applied mathematics is the dual demand it makes on its followers – competence in mathematical methods and an intense understanding of some field of experimental results. The applied mathematician has to give attention equally to deductive and inductive inference.”

Of Jeffreys, Bruce wrote, “A major part of Jeffreys’ legacy is that he set down in his papers all the essential ingredients of his argument. Basic observations, statistical methods, mathematical development and the inferential process are all present.”

Bruce further wrote, “Jeffrey’s insistence that all estimates be accompanied by a statement of formal uncertainties is now much more widely adopted in geophysics than four decades ago,… .”

From my readings of many of Bruce’s papers and reports, I infer these attitudes to have guided Bruce’s scientific work.

Bruce was a member of the Applied Mathematics Department at the University of Sydney from 1954 to 1962. He started working with Bullen on seismological problems and in 1959 submitted a doctoral thesis concerning travel times of core waves.

Jeffreys had been Bullen’s doctoral supervisor and Bruce visited him at Cambridge to complete the connection.

Bruce’s seismological contributions include:

Pioneering modern seismographic network digital systems in Northern California,
Critical new inferences on deep Earth structure,
Development of strong-motion arrays,
Education,
Introduction of numerical and statistical modeling techniques in travel-time and wave analyses.

I am going to focus on the last of these to try to give a idea of what it was like collaborating with Bruce.

One topic to which Bruce returned throughout his career was that of dispersion.

Bruce has writen about the disturbance caused by a stone thrown in a pond as follows. “Let us look … at the train of water waves on the pond. A careful watch will show that those ripples with the shortest wavelengths begin to pass ahead of the waves with longer wavelengths. After a time the train sorts itself out into a procession with the shorter waves traveling farther and farther ahead of the longer waves. This sorting of waves is called dispersion, in which the wave velocity is not constant but depends upon the wavelength (or frequency) of the waves. Seismic waves have this property in common with water waves, and … dispersion of waves has been much used in determining the properties of the Earth”

I turn to some of our collaborative work on the topic, Bolt and Brillinger (1997).

In Bolt and Butcher (1960) an algorithm and accompanying computer program were developed to create curves for the speed of energy transmission through a layer of the Earth above a lower layer. The layers were characterized by the velocities of compressional and shear waves, by the thickness of the top layer and by the ratio of densities of the two layers.

The inferential problem is: given a pertinent seismic trace estimate the layer velocities, the ratio of the densities, and the thickness of the top layer. There is also a need to provide uncertainties of the estimates.

A seismogram recorded at Uppsala, Sweden from an earthquake in Siberia was employed in our research. Theory suggested an expression for the seismogram as a sum of component waves with those of longer wavelength ( i.e lower frequency) traveling faster. The speeds depended on the parameters just referred to. Statistical theory suggested a criterion, the likelihood, for matching the data to the model and the parameters could be estimated by maximizing that criterion. In the estimation procedure, frequency components were needed and these were obtained by filtering. This is like setting a radio to different stations. The model also involved the motion of the earthquake at the source. Statistical displays allowing examination of the assumptions set down. The extent to which the fitted seismogram matched the original was quite impressive.

What did this research involve? It involved, Bruce recognizing that a contemporary statistical approach might provide an automatic solution to the problem, one including estimates of the uncertainties. It involved Bruce thinking of a pertinent data set. It involved David creating a statistical model, preparing the programs and carrying out the analysis. David had to learn some seismology and Bruce some statistics.

It was fun. We got to work together. We got to talk about the problem over lunch at the Faculty Club and out on the Bay sailing. We got to develop a new method. And as you see, research isn’t all work.

Continuing I have to tell you that in a 1960 paper Bruce beat the statistical community to what has become a very important technique – robust regression. Regression involves using data to develop an expression for the expected value of one variate as a function of others, the explanatory variates. As the work of Jeffreys made clear there are many abnormal observations occurring in data sets and they need to be dealt with. Bruce extended a method that Jeffreys had developed, for the case where there were no explanatory variables, to the regression case.

It is amusing to note the title of Bruce’s Presidential Address to the Seismological Society of America in 1975, “Abnormal seismology”. In it he indicated his intention to avoid the subject of “abnormal seismologists”.

Bruce could be stubborn, often fortified by scientific principles. I have mentioned sailing. Bruce was on the Bay one day and went aground off Richmond. A motorboat came by and offered to tow him - for $25. Bruce refused the offer. He knew about tides and 45 minutes later his boat floated free.

I have brought along two mementos of my life with Bruce,

The Centennial wine bottle. I leave it here for you to admire the label.

Also I am wearing socks that Bruce and Beverley gave me for my birthday last October. They have a map of the London Underground on them and the point of the gift was that the Bolts knew of Lorie’s and my affections for London.

Those of us who have known Bruce and his family have been privileged.

In life he made a difference. And he stimulated us to do the same. His memory will surely continue to have that effect.

Bolt, B. A. and Brillinger, D. R. (1997). Maximum likelihood solutions for layer parameters based on dynamic surface wave spectra. Physics of the Earth and Planetary Interiors 103, 337-342.

Bruce Bolt | Tributes |

Berkeley Seismological Lab
215 McCone Hall, UC Berkeley
Berkeley CA 94720-4760
www@seismo.berkeley.edu