Jim Sanderson has had a fascinating professional life. He was my PhD student in math at the University of New Mexico in the 1970s. He spent almost 20 years as a computational scientist at Los Alamos National Laboratory, working on the lab’s supercomputers. He then developed an interest in ecology, went back to school, and is now the world’s leading authority on the preservation of small wild cats around the world.
One of the great success stories of modern numerical linear algebra is the QR algorithm with the Francis shift for computing all the eigenvalues of a symmetric tridiagonal matrix. Jim Wilkinson showed that the algorithm was superior to all its competitors in the 1960s. So it was a key component of the Wilkinson and Reinsch Handbook, of EISPACK, and ultimately of the first MATLAB. It survives today as one of the primary routines in LAPACK and MATLAB.
Wilkinson established the algorithm’s global convergence and its asymptotic cubic convergence rate. But, curiously, he never analyzed its roundoff error behavior. This was Jim Sanderson’s PhD thesis. At the time Jim was writing his thesis, the academic computer science community was beginning to be interested in the notion of proving algorithms correct. Jim found it necessary to reorder the operations in Wilkinson’s implementation, but with this modification he was able to prove that the algorithm was correct. It was guaranteed to be globally, and cubically, convergent even in the presence of floating point roundoff errors.
After grad school, Jim joined the staff of the Los Alamos National Laboratory, located in the high desert mountains of northern New Mexico. Starting with the Manhattan Project, the Lab had originally developed the nation’s atomic and nuclear weapons. But with the limits imposed by international treaties, the Lab’s mission evolved from developing new weapons to maintaining existing stockpiles.
I’m not exactly sure what Jim did during his almost 20 years at Los Alamos. He is not allowed to tell me in much detail. At first it involved image processing. Later it involved large computer programs, very likely written in Fortran, run on whatever were the world’s fastest supercomputers of the day, that implemented partial differential equation models of complex nuclear reactions.
During his time at Los Alamos, Jim developed a hobby, nature photography. He says it was esoteric, artistic stuff, with photos of fallen leaves on the ground and dark clouds in a storm.
Eliot Porter was one of the first photographers to exhibit and publish color photographs of nature. He and his family eventually established a ranch in Tesuque, New Mexico, not far from Los Alamos. Jim came to work under Eliot and travel with him on several photography expeditions. Porter had a strong influence on Jim’s interest in nature photography and conservation.
In 1995 Jim left the Labs and, in middle age, returned to undergrad studies, this time majoring in biology, again at the University of New Mexico. After graduating for the second time from UNM, Jim went on to the highly regarded graduate program in Wildlife Ecology and Conservation at the University of Florida.
After completing grad school in Ecology at UFL, Jim joined Conservation International, working as a landscape ecologist.
By this time Jim was interested in small, wild cats. Big cats — lions, tigers, cheetahs, leopards — get most of the world’s attention. But there are over two dozen smaller wild cats, most of which we’ve never heard of and many of which are endangered. These cats have been an important focus in Jim’s second career.
Jim formed the Small Cat Conservation Alliance in 2006. That organization evolved into the Small Wild Cat Conservation Foundation in 2008.
One of the first cats to get Jim’s attention was the Andean Mountain Cat. Jim was the first scientist to confirm this cat’s existence. It lives above 14,000 feet in Andes of Chile and Bolivia. Before Jim’s work it was only known as a mythical creature that threatened the villagers’ chickens.
Andean_cat_1_Jim_Sanderson.jpg, Small Wild Cat Conservation Foundation.
Jim took this portrait of the cat in 1998. He says he brought all of his nature photography skills to bear when he took this shot. The photo was featured in an ad for Canon cameras that ran in the National Geographic magazine and that earned a sizable contribution to the Small Cat Conservation Alliance.
A few years ago Jim was on a long flight back from South America. He noticed a guy across the aisle doing some sort of mathematical looking puzzle in the airline magazine and asked him what it was. The guy introduced Jim (who had apparently been living away from the rest of civilization) to Sudoku. Jim returned to his seat, and to his laptop, and wrote a Fortran (more evidence of his lifestyle) program to play Sudoku.
When showed the solution to the puzzle, the guy across the aisle, not realizing that Jim had written an entire puzzle-solving program, was not impressed. “Took you a long time”, was the terse reaction. No more conversation there.
Jim told me about the incident some time later. I had a different reaction. “Time you learned to use MATLAB.”
I had been avoiding Sudoku because I was afraid I would get sucked in. To this day I’ve never done a puzzle by hand. But the program I wrote to demonstrate MATLAB to Jim is the same program that I’ve described in Cleve’s Corner, MathWorks News & Notes. I got sucked in in a different way.
Jared Diamond is a Professor of Geography and Physiology at UCLA. He is best known for the 1998 Pulitzer Prize winning book “Guns, Germs, and Steel: The Fates of Human Societies”. A documentary based on the book was produced by the National Geographic Society and broadcast on PBS.
Before writing about the great issues that made him famous, one of Diamond’s first interests in science was birds. How do various species of birds distribute themselves across strings of islands? Is the distribution random? Is it different from what it would be if the species did not interact?
In 1975, in a 102 page paper, “Assembly of species communities”, Diamond introduced the concepts of competitive exclusion and checkboard distributions. He suggests that species composition on a particular island is due to minimization of unutilized resources. The resulting distributions observed across an archipelago are alternating patterns that he called checkerboards. The data supporting his hypothesis came from his own extensive personal observations in the Bismarck Archipelago near Papua New Guinea in the western Pacific Ocean, as well as that of other visiting observers and natives.
In 1979, Edward Connor and Daniel Simberloff set off a controversy that continues to this day when they published “Assembly of species communities: chance or competition?” They pointed out that Diamond’s observed patterns had not been tested against appropriate null hypotheses. They suggested that the observed patterns could not be distinguished from patterns resulting from random distributions of the species across the subject area.
Sanderson became interested in the controversy when he read the papers while he was in school. He figured he could settle whether the patterns were random with some computer simulations. He needed to generate random matrices while maintaining constraints on the row and column sums. One approach involving the Knight’s Tour proved to be too slow to generate the $10^6$ random matrices required in this situation and faster alternatives were developed.
Jim’s simulations strongly supported Diamond’s thesis. Stuart Pimm, now a professor at Duke, had been a professor at Florida, knew Diamond, and sent Jim’s work to Diamond. A paper by the three of them, James Sanderson, Jared Diamond, and Stuart Pimm, “Pairwise Co-existence of Bismarck and Solomon Landbird Species”, was finally published in 2009. The paper criticized Connor and Simberloff’s methodology and conclusions in no uncertain terms.
Connor and Simberloff, together with Michael Collins, responded with “The checkered history of checkerboard distributions” in 2013. They claim that “Few, if any, ‘true checkerboards’ exist in these archipelagoes that could possibly have been influenced by competitive interactions.”
A pair of back-to-back notes published in December 2015 continue the interchange. One is by Diamond, Pimm, and Sanderson. “The checkered history of checkerboard distributions: comment”. The follow-up comes from Connor, Collins, and Simberloff. “The checkered history of checkerboard distributions: reply”. One of the arguments is over convex hulls on maps.
Jim and Stuart Pimm have written a book describing the whole affair, at least as of last year, that is accessible to a wider audience.
James G. Sanderson and Stuart L. Pimm, _Patterns in Nature_, The Analysis of Species Co-occurrences, The University of Chicago Press, 205 pages, 2015.
In the preface to the book they say, “Some of what happened thereafter was simply ugly, involving one of the most violent clashes of personalities in recent ecological history.” They quote Professor R. J. Putman of the University of Glasgow, even as far back as 1994, that the debate was “almost unprecedented in the apparent entrenchment and hostility of the opposing camps.”
Jim recently joined a conservation organization based in Austin that was established by some friends of his, Global Wildlife Conservation, <http://globalwildlife.org>. He is their new Director of Wild Cat Conservation.
Published with MATLAB® R2016a
Comments are closed.
3 CommentsOldest to Newest
Thanks for writing the article on Jim; it was great catching up on what he’s been doing.
After getting his PhD from UNM, Jim worked in Los Alamos with Al Carasso and me on using PDEs for deburring images. He was a coauthor of what is probably the first paper using evolutionary PDEs for image enhancement
Get the MATLAB code link broken
I’ve noticed the “Get the MATLAB code” link at the foot of Cleve’s blog posts no longer works; it doesn’t “Get the MATLAB code” when clicked.
The problem seems to have been introduced between the May 23rd “Modernization of Numerical Integration” post and the June 9th “Math and Music” post. The link works in the May 23rd post, but doesn’t work in the June 9th post.
The formatting of the “Published with MATLAB R2016a” line gives a clue. On working pages it’s displayed in a small, light gray font. On faulty pages it’s displayed as regular text.
d.write('<pre>\n'); ... d.write('</pre>\n');
with no line break in the string literals, ‘<pre>\n’ and ‘</pre>\n’.
d.write(' <pre>\n'); ... d.write('</pre> \n');
Thanks, Duncan, for letting me know about this, and for the detailed analysis. It took a while, but we traced the problem to some settings in my WordPress dashboard. I don’t know what happened to the settings last May. — Cleve