Geology and statistics are very much similar disciplines, in that both fields involve collection of data or observations and drawing inferences from them. Paul Switzer came to Stanford University in the mid-1960s and was perhaps the first person to hold a joint professorship in Geology (now Geological and Environmental Sciences) and Statistics. This is a position he has held now for 40 years, and many generations of students have benefited from his teaching. I am one of them.

Paul embodies most of the qualities one looks for in a great teacher, and certainly those are the same qualities of John Griffiths, for whom this award was named.

• He taught elementary statistics to undergraduate and graduate geologists alike, generally with no notes, no textbooks – starting with probability theory, then different forms of discontinuous and continuous distributions, followed by hypothesis testing, statistical inference, and various methods of treating multivariate data. He used a Socratic approach, with the discussion of questions and issues, resulting in conclusions that he would write in chalk on the blackboard, and the student would write down in a notebook. In effect the notebook would become the student’s textbook of statistics. Homework sets were rigorous and sometimes challenging learning experiences. Calculus had to be un-rusted on occasion (integration of probability density functions). The student came away with a healthy respect for the use and abuse of his new subject. For example, with sufficient data, one can prove nearly anything is statistically significant. I also recall being whacked on the use of the appropriate number of significant figures to state results.

• He was keenly interested in all aspects of geology, particularly the spatial sampling side. He recognized that most geologists collect samples, make interpretations, and draw conclusions from them. He tackled basic questions like the spacing between observations used to prepare geological maps, and the accuracy of the results. Early on he saw that much geological data are autocorrelated and taught the U.S. Navy how to interpolate bathymetry using what we now know as simple kriging – that was in 1964. With time, he recognized that autocorrelation processes are nonstationary in time and space. He branched out into the environmental arena and became involved with pollution issues.

• His approach to graduate study and research was pretty much the same as that in the classroom. It was very much mathematical geology and not geological mathematics. The student was forced to find his own data, formulate his problem and bring it in for discussion. A simple example was the use of measurements of lengths and widths of brachiopods as a means of separating species, a very real problem in paleontology. Another was the formulation of recurrence intervals between earthquakes as an exponential distribution. These seemed “quaint” to this economic geologist who was interested in copper and nickel distributions within magmatic ore deposits. I was in for a shock when I switched his attention to that problem.

• Discussing a problem with Paul was a voyage of discovery. The student would bring in the data, and explain the geological concepts behind it. Hypotheses were formulated, and then like a tool and die maker, Paul would select a statistical approach to prove or disprove the hypotheses. If there was nothing “in the cupboard” so to speak, this did not faze Paul. He would think about the problem and develop the tool needed. For example, we performed simple kriging on a nickel deposit. Now this deposit had a bimodal distribution for nickel, representing massive and disseminated sulfides. Our kriged estimates had a single mode, smack in the middle between the modes and not at all representative of the deposit. This led to the formulation of conditional probability distributions at sites within the deposit, and one of the first uses of indicator kriging to predict the occurrence of disseminated versus massive sulfides. This was in 1973; what is now called Geostatistics was in its infancy. There was only Matheron’s treatise at the time, and few published papers. The tools we now use like Uniform Conditioning and Multiple Indicator Kriging were ten years in the future.

• It was those voyages of discovery that were intense teaching experiences. The student was clearly out in the whitecaps, often bailing his dirty data, and there was Paul, serenely laying out in his cramped left hand the equations needed and tasks required to be completed for the next session. Occasionally the phone would ring, disturbing the peace. I remember one day. Paul used to collect convertibles, the big land yachts of the 50s and 60s, with the fender mirrors out front and the big spare tire mounted at the rear. He sold one to a guy for $400, which represented the buyer’s life savings. This was clearly an as-is-where-is transaction. Problem was, ten miles down the road, the engine blew, and the proud new owner wanted his money back. Paul really had no idea that unfortunate event would occur, but he considered a deal-was-a deal, and caveat emptor was a basic principle. As I recall, Paul promised him half his money back, and we went back to the equations.

One sample is sometimes insufficient, and I wondered whether my experience was representative. In preparing this talk I checked out a few other sources.

This from Zepu Zhang, who worked with Paul on space-time stochastic rainfall modeling from 1999 to 2004:

“I spent five happy years at Stanford, and I attribute a big part of that to the good adviser-advisee relationship. In our weekly meetings, we would immediately start detailed discussions about the model I was working on. Paul really knew what I was doing. He would have good suggestions every now and then, sometimes impromptu, some other times out of thinking on his own time. Other than occasional brief exchanges of laughter, we were very focused. There were never errands unrelated to my dissertation. [He must have discontinued the used-car business.]

“Paul has a high standard for both the idea and the exposition of the work. He did not push; instead, he let me explore freely, assured that he was always there for help. At the end of the journey, almost all the materials that went into my dissertation were developed in my final year, despite some nice ideas early on. Paul made his confidence known after I had demonstrated necessary research capabilities, which indeed is the hallmark of a great teacher.

“In the year following my graduation, we worked on publishing. Paul was involved as before. He did careful editing as a help for me from the viewpoint of a more experienced researcher, and told me this intention in a respectful manner.”

This from Ed Isaaks, who coauthored An Introduction to Applied Geostatistics in the 1980s:

“Whenever I had a problem that was really bothering me, I used to go visit Paul. He was always available, and helpful. Explanations were lucid and jargon-free.”

Finally, this from Pierre Delfiner, a continuous friend of Paul’s from the 1970s to the present and coauthor of Geostatistics: Modeling Spatial Uncertainty:

“The thing that comes to my mind when I think of Paul is his clear mind. He has a unique way of digesting a problem and reformulating it in such simplicity and clarity that it shows you the light, and you wonder why you hadn’t thought about it in this way yourself. In this respect he is the ideal sparring partner to bounce ideas off. With Matheron gone, Paul is the only person whose statistical judgment I fully trust. And since Paul is a very accessible person, the interaction with him is easy and pleasant.

“Paul is a perfectionist. I remember he once taught a class in nonparametric statistics. He advanced the field somewhat by creating new tests (I remember one for spatial independence of data), but he never published his course into a book. I asked him why; his answer was that it was not polished enough.

“As you know Paul had a joint appointment with the Departments of Statistics and Geology. Nowadays statisticians are well aware that to be useful they have to become knowledgeable in some applied discipline, but back in the seventies that was not so obvious. Paul was a precursor. Paul also played an important role, along with Geof Watson at Princeton, to introduce and promote Geostatistics within the academic world in the United States. The first time I met him was at Frascati where he presented a paper on lognormal block grade estimation.

“Paul is also an aesthete. He likes beautiful things, in particular homes, furniture, elegant proofs. He is extremely cultivated, with an interest in history and in different cultures.”

In summing up, Paul has been a model teacher in an academic milieu which these days seems overly wrapped up in funding grants and publications. Like John Griffiths, Paul’s legacy will likely be in the work of his students, and through their many contributions to the application of statistics to solving geological and environmental problems. It is with heartfelt thanks that I nominated Paul for this award, and I am pleased that the committee on awards has confirmed the nomination.

*Harry Parker*