About Tom
Thomas Heaton is a Professor Emeritus of Engineering Seismology at Caltech. Tom received a B.S. in Physics from Indiana Univ. in 1972 and a PhD in Geophysics from Caltech in 1979. He was a Senior Seismologist with Dames and Moore in 1979, after which he was appointed as a Research Geophysicist in the Pasadena Field Office of the USGS. He served as the Scientist in Charge of that office from 1985 through 1992. He next served as the USGS Project Chief of the Southern California Seismic Network from 1992 through 1995. In 1995, Caltech appointed him to be a Professor of Engineering Seismology with a joint appointment in the Division of Engineering and Applied Science and in the Division of Geological and Planetary Sciences. He became the Director of Caltech's Earthquake Engineering Research Laboratory in 2008. In his 50-year career, Tom focused his research on understanding earthquakes and developing methodologies to mitigate their impact on society. Tom says that it was his good fortune to have found key talented collaborators who were also valued friends. The following paragraphs describe many of those collaborations in roughly chronological order.
In the 1970s, Tom worked with his graduate research advisor, Donald Helmberger, to develop numerical simulations of ground motions that are expected in the near-source region of earthquake faulting on geometrically complex rupture planes. In the 1980s, Tom primarily collaborated with Stephen Hartzell (USGS, Pasadena). They developed an inversion methodology to find finite-fault rupture models that simulated ground motions recorded in both the near-source region as well as teleseismically. They wrote a series of papers that applied the methodology to records from several West Coast earthquakes.
In 1980, the US Nuclear Regulatory Agency requested that Tom assist in the review of a Nuclear Power station that was being developed in Satsop, WA. Tom collaborated with Hiroo Kanamori to study the seismic potential of the Cascadia Subduction Zone (CSZ), which at the time was largely viewed to be aseismic. Heaton and Kanamori concluded that available evidence pointed to the possibility that the CSZ is similar to other subduction zones that experienced infrequent giant earthquakes (e.g., southern Chile). In response, Heaton and Hartzell collaborated on a series of papers that were intended to estimate the nature of ground shaking that could be expected from a hypothetical giant earthquake on the CSZ. Brian Atwater, another of Tom's USGS colleagues, found sedimentary evidence of Holocene giant earthquakes, and in the 1990s, Atwater used Japanese historical records to conclude that a mysterious "orphaned" tsunami in 1700 was caused by the last giant CSZ earthquake.
In the 1980s, Tom was the sole author of two papers that were both controversial and of lasting impact. The first paper was published in 1985 and it describes the expected performance of a Seismic Computerized Alert Network. It took another thirty years for the ShakeAlert System, which is very similar to Tom's proposed system, to become a reality. The second paper followed after Tom recognized that the prevailing theory of earthquake ruptures, geometrically-similar expanding shear cracks, was incompatible with the 3-d characteristics of ruptures that were derived from the source modeling of Hartzell and Heaton. Tom argued that ruptures were comprised of slip pulses, which are a type of soliton. He further argued that strong-velocity-weakening friction was the most plausible cause for these solitons. Tom's slip-pulse paper was very controversial, and he withdrew the manuscript from the BSSA when he refused to make changes required by the reviewers. The paper was eventually published in 1989 in PEPI; Tom considers that this may be the most important paper of his career. Discovering the properties of systems that fail with self-healing slip pulses is a challenging problem in chaotic self-organizing systems and it is the subject of much of the rest of Tom's research.
Tom became the Scientist in Charge of the USGS Pasadena Office in 1985. His duties included coordination of USGS earthquake research activities in southern California. He also supervised the USGS efforts to develop and operate the Southern California Seismic Network (SCSN), which was a 250-station telemetered network of high-gain, short-period, vertical seismometers that was jointly operated with Caltech. SCSN was primarily designed to document P-wave first arrival times, which were then used to create a catalog of earthquakes that could be used to look for spatial-temporal patterns that could be used for earthquake prediction. At that time, Tom's research focused on numerically simulating recorded ground motions. Unfortunately, the very limited dynamic range of the existing SCSN meant that almost all SCSN seismograms were overdriven.
Tom collaborated with Kanamori (Director of Caltech's Seismological Laboratory) and Joseph Steim (a recent PhD from Harvard and future CEO of Quanterra, Inc.) to develop and operate a very-high-dynamic-range broad-band seismometer at Caltech's Pasadena station (PAS). Ground motions were recorded over an unheard-of range of amplitudes and frequencies. Kanamori and Heaton advocated for the deployment of a new southern California broad-band seismic network that could be used for both basic research and to provide rapid information in earthquake crises. Tom was the lead author on the National System Science Plan (NSS), which is a USGS Circular that describes the advantages of developing a new generation of regional seismic networks. The plan included development of seismic early warning as well as the development of ShakeMaps. Unfortunately, funding for USGS earthquake-related activity was very limited and implementation of the NSS plan was not possible. In contrast, Kanamori was able to secure funding from the Whittier Foundation to deploy a 16-station prototype of the system. In 1991, Kanamori, Egill Hauksson (recently hired by Caltech to lead the operation of the SCSN), and Heaton wrote an AGU Eos article that described this prototype system.
Jim Mori became the Scientist in Charge of the USGS Pasadena Office in 1992, and Tom became the USGS Project Chief of the SCSN. Mori, Hauksson, Heaton, and Kanamori collaborated to create the Caltech/USGS Broadcast of Earthquakes project (CUBE). This project created automated data processing tools together with digital communications and user displays to rapidly disseminate information to lifeline companies in seismic crises; this preceded the development of the Internet so special tools were created to broadcast information. Although CUBE was a quantum step forward in the development of real-time seismic systems, it was also controversial since it appeared that lifeline companies, who financially supported the development, were given priority access to information from a system that was mostly supported by the Federal Government.
The debate about how to support the development of a new seismic network changed when the M6.7 Northridge earthquake struck southern California in 1994. In particular, the TriNet project (USGS, Caltech, and Calif. Geol. Survey) was funded mostly by FEMA. One of the major goals of TriNet was to use seismic data that was recorded by both the SCSN and the California Strong-Motion Instrumentation Program (CSMIP) to create ShakeMaps for use by emergency managers in both the public and private sectors. The SCSN part of TriNet was a huge change for seismology. Many Caltech and USGS (Pasadena) scientists participated in the planning and implementation of TriNet. Jim Mori (USGS) and Egill Hauksson (Caltech Research Prof.) led the efforts to deploy 240 very-high-dynamic-range stations throughout southern California. The re-invented SCSN revolutionized seismic networks and it formed the basis for the eventual development of the ShakeAlert seismic warning system. Although it is not feasible to acknowledge the efforts of the many talented scientists who made the next-generation SCSN a reality, it is appropriate to single out the dedicated leadership provided by Egill Hauksson who supervised the Caltech part of SCSN.
David Wald is another key collaborator with Tom Heaton. Dave was a Caltech graduate student who was interested in the physics of strong ground motions. Heaton and Wald used the source inversion techniques that were developed by Hartzell and Heaton to study the 1989 M6.9 Loma Prieta, the 1992 M7.2 Landers, the 1906 M7.8 San Francisco, and the 1994 M6.7 Northridge earthquakes. These studies showed the importance of rupture characteristics to determine the nature of strong ground velocity in the near-source region. Prior to these important earthquakes, peak ground acceleration (pga) was widely used to characterize strong shaking. However, high-frequency shaking (i.e., acceleration) is too complex to simulate using deterministic models. Wald joined the USGS Pasadena Office and started the development of ShakeMaps. He invented the Instrumental Intensity Scale that is based on pga for felt motions and peak ground velocity (pgv) for intense shaking.
In the mid 1990s Tom's career veered strongly when he collaborated with John Hall (Caltech Civil Engineering). Heaton, Hall, Halling, and Wald published several papers that showed that near-source directivity pulses could seriously overdrive tall steel-frame buildings as well as base-isolated buildings.
Tom joined the Caltech faculty in 1995 as a Prof. of Engineering Seismology with a joint appointment in Geophysics and Civil Engineering. Although Tom's geophysics skills were well established, he was new to civil engineering. Suddenly he was teaching engineering courses that he had never taken. While much of the physics is common between engineering and geophysics, Tom needed to acquire important engineering concepts (e.g., plasticity, ductility, theory of beams, residual stress, and the strength of materials). Fortunately, Tom's engineering colleagues (especially John Hall) tutored him about structural engineering. Tom says that these new mechanics concepts (especially residual stress) radically changed his understanding of brittle failure in the Earth's crust (i.e., earthquakes).
Tom especially liked teaching Engineering Dynamics (CE 151) and he acquired a deep understanding of linear vibrations (mode theory, impulse sources, and dissipation). Over time, Tom began to include topics in nonlinear dynamics, which included dynamic chaos and self-organizing systems. Tom realized that earthquake physics is a clear example of a chaotic self-organizing system and that concepts currently used in probabilistic seismic hazards are inadequate to describe seismic risk for a system that is inherently fractal.
Brad Aagaard was the first graduate student who collaborated with Tom. John Hall was Brad's thesis advisor and Brad quickly learned the art of 3-d non-linear finite element analysis. Brad used available super computers to explore the distribution of strong ground motions in the near-source areas of hypothetical large crustal earthquakes. Aagaard, Hall, and Heaton showed that dangerous directivity pulses are not produced by thrust faults with rupture in the along-strike direction. They also investigated near-source motions from super-shear rupture velocities, and they showed that sub-shear rupture velocities radiate the most dangerous waves.
Brad also developed 3-d finite element codes to simulate the dynamics of shear failures in systems with a specified shear stress and a variety of friction laws. Aagaard and Heaton were especially interested in calculating the spatial-temporal distribution of frictional heat inferred by these models. They found that traditional slip-weakening friction models implied ubiquitous melting of rupture surfaces; that is, in order to have friction large enough to control dynamic rupture, the frictional energy must be larger than the radiated energy. In large earthquakes, the radiated energy is much larger than the minimum melting energy along the fault surface. Further, slip-weakening friction did not produce the type of slip pulses that were observed in earthquakes.
Brad and Tom used the finite-element modeling to investigate Tom's 1990 conjecture that strong-velocity-weakening friction could produce slip pulses that did not melt the rupture surface. They found that they could produce slip pulses, but when the frictional weakening was strong enough to eliminate melting, the pulses became dynamically unstable. Furthermore, the number of computations required grew as the 5th power of the resolution. Aagaard and Heaton (2008) hypothesized that low-frictional heat leads to dynamic chaos, which produces fractal slip and residual stress. Brad told Tom that using finite elements to further pursue this problem looked hopeless.
Not to be deterred, Tom looked for other types of evidence that the residual stress in the crust was a type of fractal. He collaborated with Deborah Smith to construct a 3-d fractal tensor field of residual stress. They hypothesized that true stress was the sum of residual stress (a zero-mean spatial average) together with a spatially homogenous stress that increases as plate motions accumulate stress. They combined this stress model with a simple model of Cauchy strength to create synthetic catalogs of the time, location, and focal mechanism of failures in the model. They showed that this simple model produces event catalogs that are similar to actual earthquake catalogs. Assuming that the "strength" of the crust is determined by the fractal stress (averaged over the dimension of the failure plane), they showed that the strength of the crust decreases as (length) -1/4 (Smith and Heaton, 2010).
Although it seemed clear that 3-d finite element calculations of chaotic rupture were not feasible, Tom thought that simpler 1-d dynamic models of slip pulses could provide insight into fractal residual stress. Fortunately, Ahmed Elbanna found Tom's class that covered dynamic chaos to be especially stimulating. Ahmed created a simple 1-d spring-block-slider model (sbm) that also had strong velocity weakening friction. This system was simple enough that numerical calculations could be performed with enough precision to study the ensuing chaos. In particular, they showed that the system generated spatially complex slip pulses. Furthermore, after a period of start-up events, the system evolved into a residual stress state that produced power-law behavior (Gutenberg-Richter b-values, and a residual stress with power-law power spectra, that was similar to the stress model of Smith and Heaton). Tom and Ahmed were intrigued by the fact that a very large numerical calculation was required to simulate each sbm event (this is caused by the strong positive feedback between the slip and friction law). After much thought, they concluded that there should be a way to greatly simplify the calculation by tracking energy in the system (potential spring energy, kinetic energy of the blocks, and frictional dissipation) as a slip pulse propagates through the system. They derived the "pulse energy equation," which is very different from other equations in mechanics (Elbanna and Heaton, 2012). The pulse-energy equation is much easier to solve than the full sbm system and it produces the same self-organization characteristics. Tom and Ahmed are very excited by this development, even though few other researchers seem to have noticed this work. If you are interested in knowing more about this work, Tom suggests that you read Chapter 8 of his class notes on Engineering Seismology (CE 181). This contains a summary of this topic and most of the discussion is not published elsewhere. www.its.caltech.edu/~heatont/Eng_Seism_Notes (caltech.edu)
Tom collaborated with his PhD advisee, John Clinton, to determine seismometer designs that were optimized for recording both strong shaking as well as shaking from a wide band of regional earthquakes. They produced a summary figure that shows the range of several important systems that are superimposed on the shaking expected from a wide range of seismic sources. This figure is widely used in network seismology. John and Tom also analyzed the motions of Millikan Library; the analysis includes many earthquakes, forced harmonic tests, and years of continuously recorded ambient vibrations. Clinton, Bradford, Heaton, and Favela, 2006) documented the fact that 1) the apparent natural frequency of Millikan drops with the amplitude of the vibration, 2) the fundamental frequency dropped significantly over decades, and 3) the frequency of the building increases by several percent when it rains.
Although Tom had long dreamed of creating an early warning system, he knew that it would require the development of real-time computer systems to transform seismic data streams into predictions of shaking that was about to occur. He collaborated with one of his PhD students, Georgia Cua, to develop strategies for automated alerting. They reasoned that a skilled human seismologist could be very effective, except that humans are slow and they need to sleep. They argued that we should develop a computer analysis system that mimics the judgement of humans, and they called the prototype The Virtual Seismologist. Fortunately, James Beck (Caltech CE) also collaborated on this project. Jim is one the foremost experts on Bayesian inference theory, which forms the framework of systems that find optimal solutions to problems with incomplete data and prior information. Although Cua's insights about using inference theory are considered to be the optimal framework for an alerting system, simpler solutions are used in current systems.
Up to that point, earthquake alerting algorithms assumed that earthquakes could be described as a point source. Tom recognized that the optimal benefit of alerting is in large-magnitude earthquakes that have large rupture surfaces. Tom collaborated with his PhD student, Masumi Yamada, to develop a methodology to track finite ruptures in real time. They developed a simple parameter that indicated whether a station was likely to be within 10 km of a rupture surface. They then used this function to track the temporal-spatial evolution of the rupture. They also developed a technique in 2007 to estimate the spatial distribution of slip based on peak ground displacements.
By 2007, earthquake alerting was entering a new phase, the development of a demonstration system using the SCSN. Maren Böse had just finished an innovative PhD dissertation (Univ. of Karlsruhe) on earthquake alerting and she was hired as a post-doctoral student to help in this effort. Tom and Maren collaborated on several projects which especially focused on tracking finite rupture in real time. In her PhD work, Maren devised a scheme to fit the spatial distribution of pga with precomputed spatial templates of pga for a variety of earthquake rupture lengths; she called this algorithm FinDer. Clearly, even though they had not yet met, Maren and Tom were on the same wavelength about tracking large ruptures in real time. Maren is particularly good at transforming ideas into working systems. It's harder than it sounds.
In the past decade, Tom collaborated with Sarah Minson (USGS) to create a framework to use Bayesian inference theory to combine information from diverse alerting algorithms (Minson, Wu, Beck, and Heaton, 2017). Instead of focusing on estimating source parameters, Minson argued for the development of predicted Shakemaps that evolve as new information is included. This approach has been adopted by the ShakeAlert project, but significant coding and testing is required before this new approach becomes a reality.
Lucy Yin and Becky Roh are in the group of Tom's last graduate students. They worked on practical ways to include foreshocks as prior information in earthquake alerting. They both studied machine learning at Caltech and they suggested new ways to use AI techniques to rapidly identify the characteristics of an earthquake in real time.
Tom also collaborated with Men-Andrin Meier (Caltech and ETH) on several important studies in earthquake alerting and earthquake rupture physics. They collected a large number of near-source P waves and showed that the typical moment-rate function is independent of the eventual size of an earthquake for the first half of a second. They showed that the typical moment-rate function changes in character after half a second (the duration of a typical M5). They suggested that earthquakes grow as expanding cracks until they transition into slip pulses and that this transition is at a rupture radius of about 1 km. Tom and Men-Andrin also studied the temporal structure of moment-rate functions from large earthquakes and found that their median time characteristic is approximately an isosceles triangle (even though individual time functions are extraordinarily complex).
Although Tom focused much of his attention on the development of algorithms for seismic alerting, he also collaborated on the development of working systems. Richard Allen became interested in earthquake alerting when he was a post doc at Caltech working with Hiroo Kanamori in 2003. Soon thereafter Richard joined the faculty at the Univ. of Calif. Berkeley, and he focused on developing operational earthquake alerting in the Bay Area. About that time, Heaton and Hauksson were also developing operational alerting systems at Caltech. Allen, Hauksson, and Heaton decided that coordination between the Caltech and Berkeley alerting projects was the most effective path forward. They received support from the USGS to develop a working demonstration system in 2006. The ShakeAlert demonstration system went live in 2012. Until that time, the USGS debated the future significance of earthquake alerting. However, following the 2011 Tohoku, Japan earthquake, the USGS formally committed to coordinating development of an alerting system for the US West Coast. The ShakeAlert system became operational in 2016. Although this system is the result of the dedicated work of dozens of scientists and engineers, Tom identifies Douglas Given's key role (USGS) in coordinating the management of ShakeAlert. Tom says that he could see the eventual benefits of an alerting system back in 1979, but he never imagined that it would require so much development by so many people.
In the last decade of his research career, Tom continued to work on finding the most effective strategies to design urban areas for earthquake resilience. Simultaneously, there's been a groundswell of work to develop Performance Based Earthquake Engineering (PBEE). Recognizing that it is impractical to design earthquake-proof structures, PBEE strives to design structures that can survive very infrequent shaking. For example, it's common to design high-rises such that they are expected to collapse in the motions that have a recurrence time of 2,500 years. Although this seems to be a rational approach to resilient design, Tom was skeptical that it is possible to estimate the 2,500-year motions, especially in the near-source regions of earthquakes larger than M7.
Tom worked with John Hall and Anna Olsen to simulate the collapse of code-compliant steel-frame buildings that were subjected to the simulated ground motions of large earthquakes. They used Brad Aagaard's simulations of ground motion in the M7.8 1906 San Francisco earthquake to show that severe damage and collapse could be expected in many urban regions. It seems that the long-period ground motions associated with being near an earthquake with large slip were significantly larger than was assumed in studies of probabilistic seismic hazard analysis (psha).
Olsen, Hall, and Heaton followed up by investigating whether response spectral acceleration at 5% damping is a useful characterization of ground motions that can cause collapse. They found that the combination of large pgv (> 60 cm/s) and large pgd (> 60 cm) is especially dangerous for high-rise buildings.
Shiyan Song and Kenny Buyco collaborated with Tom to show that 70% damped spectral acceleration at twice the elastic period is a much better predictor of collapse than the industry standard of 5% damping at the elastic period. Buyco, Roh, and Heaton also showed that standard processing of near-source accelerograms produces ground motion estimates that can decrease the implied impact of a record for tall buildings.
Tom's experience as a seismologist told him that estimating the statistics of fault slip is key to estimating the shaking that is dangerous to flexible buildings. Based on his work with rupture physics, Tom knew that we are dealing with a chaotic system that is described by power-law statistics (aka, Pareto distributions). Olsen, Yamada, and Heaton showed that long-period near-fault motions cannot be estimated using currently available data; the median motions increase as more data is added. Tom has argued that it is not possible to calculate the exceedance statistics of long-period ground motions (we have little data about the things that matter most). He argues that the objective of earthquake engineering is to identify the designs that are most robust; that is, use the building designs that are least vulnerable to unknown long-period shaking.
Through his career, Tom discovered the importance of seismic networks to drive understanding and innovation. In about 2010, MEMS accelerometers with sufficient sensitivity for earthquake studies could be purchased. These solid-state devices were inexpensive ($35 per 3-component accelerometer), reliable, well-calibrated, and easy to install. Several seismologists recognized that these devices were a potential game-changer for strong-motion networks. At Caltech, Rob Clayton, Monica Kohler, Mani Chandy (Computer Science) and Tom collaborated on the Community Seismic Network (CSN), which was funded by the Gordon and Betty Moore Foundation. The original plan called for the installation of CSN stations that were operated by volunteers with PCs that were connected to the Internet. Although numerous CSN stations were deployed in this way, it became clear that the volunteers were the weak link in the plan (they lost interest, their PCs changed, etc.). In response, CSN focused its attention on stand-alone micro-computers that could be networked using donated Internet connections. CSN has concentrated on three projects: 1) 700 stations in campuses of the LA Unified School District, 2) 220 stations located on the JPL campus, and 3) several high-rise buildings.
Unfortunately, the network was deployed during a long drought of felt earthquakes in the Los Angeles region and the CSN developers struggled to find the support to maintain the network. However, when a M7.1 struck Ridgecrest in 2019, CSN produced hundreds of high-quality records in the Los Angeles Basin. These records reveal surprising amplification (up to ten times) for long-period motions. CSN is a new paradigm for recording strong shaking in urban areas and it promises to revolutionize engineering seismology.
In 2015, Tom was contacted by Shervin Taghavi (Caltech PhD in Electrical Engineering) about potential applications of optical systems to measure motions. Tom had been intrigued by the notion that building motions could be derived from videos that showed motions. To test the idea, Taghavi recorded the motions of the deck of the Vincent Thomas Bridge. Videos showed clearly-resolvable motions caused by truck traffic. Tom and Shervin realized that video monitoring of bridges could be used to weigh the traffic. They developed a system to detect overweight trucks using telemetered video monitoring. They have started a company (WeighCam) to provide this information to agencies that maintain highways. Taghavi, Heaton, and Caltech are joint owners of a patent for this promising new technology.
Tom is famous for asking difficult questions. Perhaps this is because he reported a statistically significant correlation between solid-body tidal stress and earthquake occurrence in his first noteworthy paper in 1975. After that work, he collected data for another 6 years and used it to rerun his statistical tests. In 1982, he reported that the statistical test was not repeatable, and he concluded that it was invalid to calculate statistical significance about a pattern that was suggested by inspection of the original data set. Ironically, the incorrect paper from 1975 continues to be cited at a higher rate than the correct paper from 1982.
Personal
Tom was born in Lawrence, Kansas in 1951 where his father, Robert Heaton, studied mathematics. In 1956, his father joined the Mathematics faculty at Rutgers Univ. and Tom's childhood was spent in Piscataway, NJ. Although Tom showed talent when it came to science, he was a mediocre student when it came to anything involving memorization (spelling, reading, etc.). Many years later, his colleague, Ralph Wiggins, told Tom that he was clearly dyslexic. Tom is famously absent minded; for example, he has trouble remembering names.
Tom's mother, Jacquelin Heaton, had an aptitude for music and she eventually became the Music Director of the Piscataway school system. Tom also has a lifelong love of music, and he is a talented guitarist. He owns numerous acoustic and electric guitars and he records music in his home studio. He was also in a band for many years.
Tom met Norma Hart Graham in Honors Linear Algebra class at Indiana Univ. in 1970. Norma was a business student and Tom assisted her by helping with her homework. Of course, Norma got an A, while Tom got a B. Fortunately, Tom excelled in his physics classes, and he was accepted by Caltech for graduate work in geophysics. At the same time, Norma graduated at the top of her class in the IU Business School. Simultaneously, Norma was granted a divorce from her first marriage. Tom and Norma began dating; they were soon engaged, and Norma decided to move to Pasadena with her three-year-old son, Adam Graham. What could possibly go wrong? Apparently not much since they are still married 49 years later. Norma became a CPA in 1975 and she gave birth to Emily and Wyllis Heaton while Tom was still a grad student (times have changed). Adam (wife and three children) and Wyllis (wife) live in Montecito, CA, and Emily (husband and two children) lives in Santa Rosa, CA. Norma spent 7 years as the CFO of Solheim Lutheran Home (a senior living community in Eagle Rock with a capacity of 190) and 16 years as its Executive Director.
In 2004, Tom was diagnosed with Fuch's dystrophy, which is a genetic disease of the corneas. By 2007, Tom was becoming functionally blind. He received a cornea transplant in his left eye, which was a failure. Fortunately, the second left-eye transplant in 2008 was a success. Cornea transplants were then attempted for the right eye. None of these were successful and Tom lost total function in his right eye after a severe infection in 2009. Tom continues to function with the use of his left eye, but his overall visual acuity is poor. If he fails to greet you when you see him, it is likely that he hasn't seen you clearly enough to recognize you.
Tom has always been interested in sports cars (especially red convertibles). He enjoyed being a motorcyclist for 50 years, but he sold his last treasured motorcycle when he turned 70. He continues to enjoy hiking and bicycling.
Tom identifies acceptance into the Caltech community as a great gift. Upon their return from a scientific meeting, Hiroo Kanamori told Tom that "it feels like home." Tom agrees.
