Thomas H. Heaton |
|
Emeritus Professor of Geophysics and Civil Engineering Ph.D., 1978, Geophysics, California Institute of Technology B.S., 1972, Physics, Indiana University email: heaton@caltech.edu |
|
| Strong Ground Motion Research |
I have included most of my knowledge of earthquakes and buildings in an on-line textbook that I wrote over several decades of teaching Engineering Seismology at Caltech, Physics of Earthquakes and Buildings can be down loaded. Although smaller earthquakes are far more numerous, large earthquakes (M > 7.5) account for most of the slip in plate tectonics. If we assume that M 8.0 is the largest earthquake magnitude that an earthquake can have in California, then there is three times as much radiated energy in the M 7 to 8 earthquakes as there is in all other earthquakes smaller than M 7. Although large earthquakes are infrequent, they are inevitable since they are the major actors in plate tectonics. What will happen when a large magnitude earthquake attacks one of our cities? The answer depends on the particulars of the earthquake and the capacity of the buildings that are shaken. Even if we knew the magnitude and location of a future earthquake, it's still hard to predict the shaking; ground shaking often varies by a factor of ten between sites that are equidistant from an earthquake. To deal with the large variability in observed shaking, it has become popular to construct probabilistic models of exceeding a given intensity of shaking (Probabilistic Seismic Hazard Analysis, PSHA). PSHA models are constructed using statistical analysis of records of strong shaking. A key questions is how well these records inform us about what will happen in future earthquakes. There are currently enough records to characterize motions that can occur in earthquakes up to about M 7. However, larger earthquakes are so infrequent that there are too few records to characterize the range of shaking that occurs in great earthquakes (e.g., the 1906 San Francisco earthquake). Therefore PSHA must extrapolate relationships between ground motion and earthquake magnitude to large magnitudes. As currently practiced, most of these extrapolations are based on log-normal statistical models (Gaussian distributions) that are commonly used in actuarial science. While this type of statistics may be appropriate for high-frequency ground shaking (peak ground acceleration), the statistics of low-frequency ground motions are best described with heavy-tailed power laws (sometimes called a Pareto Distribution). Unfortunately, the current practice of using log-normal statistics seriously underestimates the size of long-period ground motions that will occur in future earthquakes. My research is in both earth sciences (understanding the physics of shaking) and in earthquake engineering (understanding the physics of yielding buildings). There is inconsistency between earth scientists and earthquake engineers about the significance of large magnitude earthquakes. Much of my work focused on a more complete understanding of the nature of ground shaking close to large earthquakes. That is, ground motions from large earthquakes are simulated by propagating waves through 3-dimensional earth structure models. The models produce realistic estimates of the large displacementsthat occur in great earthquakes (several meters in several seconds) . While accelerations that are associated with these large displacements may not be large enough to cause failure of strong, shear-wall buildings (most of California's construction of 3 stories and less), they may cause severe deformations and collapse for flexible buildings (almost all buildings taller than 8 stories are flexible) that rely heavily on ductility for their performance in large earthquakes. I have collaborated with John F. Hall, who is a Caltech emeritus Professor of Civil Engineering.
We (Jing Yang) also investigated the potential performance of steel moment-resisting-frame buildings in large subduction zone earthquakes. We have simulated the deformations and damage that would have occurred to such buildings in the M 8.3 Tokachi-Oki earthquake (2003). Although there we no such buildings present on the island of Hokkaido during this earthquake, there were 275 strong motion records which we used as the basis of our study. In addition, we used this data as the basis of an empirical Green’s function study of the potential effects of a giant (M>9) subduction earthquakes on high-rise buildings in the cities of Seattle, Portland, and Vancouver. Our simulations indicate that long-period shaking from a giant Cascadia earthquake will last for three to five minutes. Furthermore, long-period period (2 to 8 sec.) ground motions will be strongly amplified in the Seattle basin. Because the down-dip extent of rupture on the subduction zone is unknown, we simulated motions for three different cases: 1) the rupture is confined to the offshore part of the zone, 2) rupture extends about 20 km east of the coast, and 3) rupture extends to the eastern margin of the Olympic peninsula. Shaking from cases 2 and 3 is strong enough to induce large nonlinear deformations for building simulations in the Seattle basin. In many of the simulations, collapse is indicated. It seems clear that current design procedures for Seattle high-rises do not assure collapse prevention in the case of a giant Cascadia earthquake. This work is described in Jing Yang's Ph.D. dissertation (pdf). Anna Olsen and I studied the performance of steel moment-resisting-frame buildings and base-isolated buildings in simulations of large crustal earthquakes in California. These include simulations of the 1906 San Francisco earthquake (collaboration with Brad Aagaard) and simulations of several plausible earthquakes in the Los Angeles Basin. We showed that a repeat of the 1906 earthquake may cause irreparable damage or collapse of many tall buildings in San Francisco. The collapse hazard is about five times higher for steel frame buildings with brittle welds (pdf). This includes most buildings constructed before the 1994 Northridge earthquake, which revealed this flaw in building construction. Although traditional response spectra are best for predicting building deformation for moderate shaking, our analysis indicates that a combination of peak ground velocity and peak ground displacement is actually a better predictor of building collapse. We also showed that base-isolated buildings that are typical of the current state of the art are likely to experience violent impacts with their foundations for many sites within 10 km of the San Andreas fault. We (Masumi Yamada and Anna Olsen) showed that the seismic hazard from long-period near-source ground motions is fundamentally different from the seismic hazard from short-period motions (pdf). While the hazard from near-source short-period motions can be well characterized by a log-normal distribution, the hazard from near-source long-period motions seems best described by a log-uniform distribution, which is a type of heavy-tailed Pareto distribution. Unfortunately the uncertainties introduced by such a distribution are very high. Current procedures in performance based earthquake engineering may not appropriately capture the risk associated with large earthquakes and long-period buildings. The PhD work of Shiyan Song shows that the current practice of characterizing the intensity of ground shaking with 5% damped response spectral acceleration at the elastic period of a building does a poor job of predicting collapse of buildings when compared to non-linear finite element analysis of steel special-moment-resisting-frame buildings. A better approach to equivalent linear analysis is to assume a free period that is lengthened by square root of the ductility, and damping that is on the order of 70% of critical. That is, buildings that are undergoing large plastic strains are very poor resonators. We showed that when the 70%-damped spectral acceleration at 3/2 the elastic period exceeds the pushover yield strength of a building, then the building is likely to collapse. We show how this measure of ground motion is related to peak ground velocity and peak ground displacement that is used by Olsen, Hall, and Heaton. Kenny Buyco has extended the work of Song by comparing the ability of heavily-damped response spectra to estimate the drifts in a variety of steel moment resisting frame buildings. Buyco has consttructed finite-element models of several simple steel-frame buildings that meet current and past code. He has performed incremental dynamic analysis of these buildings so that we can document how code changes have affected the collapse resistance as a functon of time. Buyco's work documents important inconsistencies in the way in which spectrum-compatible ground motions are currently used for performance-based structural design. Base-Isolated BuildingsIt seems that seismic isolation systems are becoming very popular for buildings in seismic regions. While the potential benefits of isolation systems are well documented (reduced inertial forces for the structure and contents of a building), it seems that very little attention is given to the fact that base-isolated buildings may perform poorly in the largest earthquakes. That is, there is a maximum displacement that can be accommodated by isolators. These displacement maxima are determined by a number of factors, but especially the diameter of the isolator. Many recent designs have maximum isolator displacements of 40 to 50 cm. If the ground moves moves more than the allowable isolator displacement, and if the motion occurs fast enough, then the building may experience an impact between its structure and a concrete wall in the foundation. Such impacts are likely to damage the structure of the building and to cause very high acceleration transients within the building. For example, the International Terminal of San Francisco International Airport is base isolated and it has a maximum isolator displacement of 40 cm and a 3 sec. equivalent period of the isolators. The structure is located about 4 km from the San Andreas fault that experienced 3.5 m of slip in the 1906 earthquake. Simulations of ground motions for the 1906 earthquake produce ground motions at SFO that clearly exceed the design for this award-winning structure (pdf). Amazingly, it is almost never mentioned that a repeat of the 1906 earthquake would probably cause severe damage to this structure. The San Bernardino Justice Center is another example of an award-winning base-isolated structure. In this case, 5-sec triple pendulum isolators are employed to isolate this 11-story building that is located 8 km from the San Andreas fault. The maximum isolator displacement of 110 cm may be too small if this fault experiences large slip in a future large earthquake (most earth scientists consider a large southern San Andreas earthquake to be likely within the next century; the ShakeOut Scenario assumes slips as large as 12 m on some fault segments). Again, available documents only discuss the benefits of the isolation system, and the possibility that the system will result in a violent collision between the building and its foundation is not mentioned. It appears to me that something is seriously amiss with the current situation. Structural engineers are advising clients that installation of seismic isolators will greatly decrease their vulnerability, even though it seems clear that some base-isolated structures are destined to experience violent impacts in future large earthquakes. I suspect that much of the miscommunication between structural engineers and the earth scientists is the result of the National Seismic Hazard maps produced by the USGS. These maps provide estimates of peak shaking intensity for specified time periods (e.g. 10% in 50 years), where the shaking intensity is described with either peak ground acceleration (pga), or response spectral acceleration (sa) at a variety periods of 0.2 s out to 10s. The National Hazrd maps are constructed from models that describe the average repeat time for earthquakes on segments of known active faults. The earthquakes that correspond to these segments are assigned a magnitude (this magnitude depends on the assumed length of the rupture). For each potential earthquake a ground motion prediction equation is used to assign shaking intensity for every point on a regional maps. Unfortunately, none of these shaking intensity parameters provide any information about the maximum displacement required of a base isolator. That is best described by response spectral displacement in the period band from 2 to 5 seconds. This procedure assumes that near-source long period motions are well characterized by the earthquake magnitude and the appropriate gmpe, which in the case of the National Hazards Maps, is a complex blend of five separate gmpe's known as NGA 2. NGA 2 gmpe's are obtained by least-quares regressions of strong motion records that are contained in the PEER database. One critical problem is that there are relatively few near-source records in this database. Furthermore, strong shaking close to a large earthquake can often cause soils to compact. This compaction can cause local tilts at the strong motion accelerograph. A change in tilt introduces a constant bias to a record. Double integration of an acceleration record that contains a change in tilt typically produces displacements that are quadratic in time. In order to avoid quadratic displacements, the records in the PEER database are filtered with a high-pass Butterworth filter. Although this process removes most of the effects of tilt, it also removes important parts of the true displacement. In reality, the near-source long-period motions are best derived from the slip distribution on a fault and the continuum equations. Most seismic hazard calculations assume that shaking statistics are described by log-normal variations about mean values that are simple functions of earthquake magnitude and the distance between the site and a potential rupture. Unfortunately, the statistics of long-period ground motions are definitely not log-normal; they are best described as a Pareto Distribution in which infrequent events control the exceedence probabilities that are important for base-isolated buildings (see Yamada, Olsen, and Heaton). This is a critically important issue; we are commonly told that these long-period structures are designed for the 2,500-yr ground motion, when in fact, large earthquakes that will occur in the next several centuries are likely to cause severe damage of these important structures. |
| Earthquake Rupture Physics and Crustal Stress |
Much of the deformation of the Earth's crust occurs as earthquake rupture. Therefore, it is of critical importance to understand the fundamental dynamics of earthquake rupture to understand the stress state of the crust. A short description of the problem can be found at Live Science. We are particularly interested in understanding the origins of spatially heterogeneous slip in earthquakes. There is compelling evidence that slip in earthquakes and stress in the Earth’s crust are spatially heterogeneous, and perhaps fractal. We have been pursuing two different approaches to understand the dynamic properties of this system. The first approach is a long-standing collaboration with Dr. Brad Aagaard (USGS, Menlo Park) and it consists of constructing 3-dimensionional finite-element models of the Earth's crust, which are controlled by dynamic friction on fault planes. The models include the effects of gravity so that crustal stresses are consistent with the topography of the Earth's surface and density variations in the crust. The models allow us to follow the partitioning of elastic and gravitational potential energy into radiated seismic waves, fracture energy, and frictional heating on faults. Using estimates or bounds on wave energy, fracture energy, and heat energy, it is possible to put bounds on crustal deviatoric stress. Despite steady progress in simulating dynamic earthquake ruptures, there are limitations of this approach to understanding the dynamic properties of the crust. In particular, recent experiments in dynamic friction suggest that there are rapid transitions between high static friction (>200 MPa at 10 km depth) and very low dynamic friction (<5 MPa). These strong transitions in friction point to very localized slip pulses that propagate unsteadily along faults. Unfortunately, simulation of dynamic rupture with these friction laws requires enormous spatial grids with very fine time resolution. We (Jing Liu-Zeng) have constructed fractal models of slip that are compatible with observations of slip vs. rupture length scaling and also with earthquake frequency vs. magnitude statistics. In addition we (Deborah Smith) have constructed a 3-dimensional fractal model of tensor stress that we use to simulate catalogs of earthquake locations and focal mechanisms. This model predicts that traditional inversions of focal mechanism catalogs for average stress orientation may provide results that are seriously biased towards the orientation of the stress rate function. It also predicts that the strength of the crust depends on the length scale over which failure occurs. We (Ahmed Elbanna and I) investigated the statistical relationship between fractal stress and fractal slip. Ahmed Elbanna, Nadia Lapusta, and I have been able to show that there are no steady-state solutions to the problem of a propagating slip pulse that is the result of pure rate-weakening friction. Unfortunately, it is not technically feasible to numerically simulate the long-time behavior of a sliding surface subject to pure rate-weakening friction. In order to gain insight into this challenging dynamics problem, Ahmed Elbanna and I studied the behavior of pure rate weakening friction for a simple spring-block-slider model. We found that this system is inherently a chaotic system that self organizes into complex prestress states that appear to be fractal in nature (wavenumber spectra described by power laws). We show that any rational definition of the material strength of this system depends on the length scale over which the strength is measured. In particular, larger systems operate at smaller average stresses than smaller systems. Furthermore, the scale dependence of the strength is related to the b-value of the seismic events in the system. b-values approaching 1.5 correspond to an interface that slips during many small magnitude events. On the other hand, b-values less than 0.5 correspond to systems that primarily slip in large events. We call the small b-value systems, "brittle" and they have strong length scale dependence with increasing size. The 1.5 b-value systems are more ductile and they have weak length scale dependence of strength. You can read about these interesting findings in Elbanna's PhD thesis pdf. Perhaps the most exciting development of our work on chaotic spring-block sliders is the discovery of the Pulse-Energy Equation pdf. Full numerical solution of the spring-block-slider model requires very large computations. We derived a simple 1-d ordinary differential equation that tracks energy changes in a system as a slip pulse propagates along the interface. This is a new class of equation that closely mimics the full numerical solution. Furthermore, it seems to work over an extreme range of time and spatial scales. Best of all, when it is run over a long time scale (many events), it self organizes into a state that is similar to the state that is the result of the full numerical solution. I view discovery of this equation as a fundamental breakthrough. |
| Earthquake Warning Systems |
I have had a decades-long interest in developing automated systems to utilize seismic data to help society to respond during an earthquake crisis. In 1985, I published "A model for a seismic computerized alert network," which described a framework for developing a system to alert users of the arrival time and shaking amplitude of seismic waves that were propagating towards a particular user. That paper generated widespread interest and it was the basis for many of the design requirements for the Southern California Seismic Network (SCSN). Through the years, I have collaborated with numerous colleagues to try to make the seismic computerized alert network (SCAN) a reality. I worked with Jim Mori (USGS), Hiroo Kanamori (Caltech), and Egill Hauksson (Caltech) to design and deploy the Southern California Seismic Network (SCSN). I also worked with David Wald (USGS) to develop the first versions of ShakeMap, which is now an important emergency management tool that is supported by the NEIC of the USGS. More recently, I have collaborated with Richard Allen (UCB) and Egill Hauksson (Caltech) to develop CISN ShakeAlert, which is currently an operating demonstration system that uses CISN data to alert dozens of test users about shaking that they are about to experience. My role in ShakeAlert development has focused on the development of new automated algorithms that analyze data and then predict the characteristics of impending shaking. I worked with Georgia Cua to develop a Bayesian statistical framework that incorporates prior information with real-time seismic information to produce a probabilistic model of future shaking. This work was the basis of the Virtual Seismologist (VS) framework that is currently being developed corroboratively with ETH, Zurich (John Clinton and Yannik Behr). The Virtual Seismologist (VS method) is based on the type of robust analysis that a human would perform if they had the time. We use envelopes of acceleration, velocity, and displacement as the basic data input to a Bayesian framework that also incorporates other types of information (e.g., topology of the seismic network, recent seismic activity). I am working with Lucy Yin to incorporate current seismicity information (e.g., potential foreshocks) into the analysis. I am also working with Gokcan Karakus on developing an algorithm that checks the match between envelopes of recorded strong motion data and predictions that result from the alerting system. This promises to significantly improve the reliability of the system. I have also been working with Men Andrin-Meier (ETH) to develop a Bayesian inference algorithm (the Gutenberg Algorithm) to statistically analyze the output of real-time filter banks that are an efficient type of wavelet transform. The Gutenberg Algorithm promises to make seismic alerting systems significantly faster than existing algorithms; the filter banks deliver the vital information with delays that approach the theoretical minimum. Most existing algorithms are designed to determine the location and magnitude of a point source that best explains available data. However, in earthquakes larger than M 7, it is critical to know the spatial extent of a long rupture. We are also working on methodologies that will provide real-time estimates of rupture geometry and fault slip. Masumi Yamada and I extended the VS framework to include the analysis of time-evolving finite ruptures. We developed algorithms that determine whether or not a station is located in the near-source region of a rupture (pdf). This work was extended in a collaboration with Maren Bose to develop the FindEr algorithm (pdf) that uses peak acceleration data to track the propagation of long ruptures. Maren Bose and I also developed a real-time algorithm that estimates the spatio-temporal distribution of slip by projection of peak ground displacement data (GPSlip). This work was based on research with Masumi Yamada. One of the challenges of seismic alerting is to estimate the probability of future shaking for earthquakes that are still propagating. More recently, I have been collaborating with Maren Bose and Sarah Minson (USGS) to develop a Bayesian inference framework to combine the information that is coming from numerous algorithms. This is a challenging problem in which we must determine which information is significant and independent of other information. This framework is currently being developed into C++ computer code called Finder-BEFORES. It's one thing to provide rapid alerting information and it's quite another thing to use the information to make the best decisions in an ongoing emergency (kind of like battlefield management). I have been collaborating with James Beck and Stephen Wu to develop a framework (ePAD) for deciding when to trigger automated actions based on stream of information broadcast by the ShakeAlert system. For example, some actions may be costly to implement and it may be appropriate to delay an action until enough data is analyzed to insure that the ground motion prediction is reliable. Elevator control is an obvious candidate for earthquake alerting. The motions of tall buildings are often very different from the motions at the base of a building. For example, tall buildings (20 stories) in Mexico City resonated sympathetically with resonance of the lake beds underlying Mexico City in the 1985 Michoacan earthquake. People on the ground adjacent to buildings only experienced very mild shaking while nearby tall buildings swayed violently and often collapsed. The combination of a large magnitude event located some 200 miles from Mexico City would have allowed a prediction that tall buildings would sway for over a minute in this event. Unfortunately, earthquake alerting would probably not have saved people who were crushed by the collapsing buildings, a system could predict that the buildings would experience a long duration of swaying motion. Ming Hei Cheng and I have proposed a methodology to predict the nature of shaking in tall buildings based on the known resonant period of a building, the location of the building relative to the earthquake, and the location of a person in the building (pdf). Clearly, some day in the future, residents of tall buildings will be told by their smart phones about the characteristics of the shaking that they are about to experience. Caltech, the Univ. of Calif. at Berkeley (UCB), and Univ. of Washington (UW), the USGS, the California Office of Emergency Services, and ETH Zurich are actively developing the ShakeAlert system (pdf), which is a working demonstration project to develop earthquake early warning on the West Coast of the US. This system has been funded by the U.S. Geological Survey and the Gordon and Betty Moore Foundation. More information about Caltech's role can be found on our Caltech earthquake alerting website. |
Studies of Building Vibrations |
The overall goal of this research is to develop tools and a framework to describe the structural properties of buildings. In particular, there is growing interest in developing a building rating system. That is, interested individuals would be able to discover pertinent information about the structural integrity of individual buildings so that they could make more informed decisions about whether of not to occupy that building. While knowledge of the building's structural design is critical to understanding its seismic resistance (see the PhD research of Kenny Buyco), it is also of critical importance that we are able to characterize and monitor the vibrational characteristics of individual buildings. This means that we must extend our seismic networks so that they also document the vibrations of buildings. We (Monica Kohler) are investigating the vibrations of buildings that are excited by a wide number of sources, including wind, explosions, machinery, and earthquakes of all sizes. We have installed an advanced 24-bit, 140 dB seismic station that continuously records the 9-story Millikan Library on the Caltech campus, a building which has been the source of several interesting mysteries. For example, when the building's fundamental modes (north-south, east-west, and torsion) are excited by a 1-hp eccentric shaker operated on the building's roof, harmonic seismic waves are observed at the building's eigen-frequencies throughout the Pasadena area; they can even be detected on seismometers just north of the US-Mexican border, which is about 250 km away (see Javier Favela’s dissertation). Another interesting mystery of Millikan Library is the fact that the natural frequencies of the fundamental modes (north-south, east-west, and torsion) all increase by several percent just following significant rain storms. These increases in frequency slowly decrease over a period of several days. We have been using advanced time-frequency representations (the Wigner-Ville distribution) to investigate how these natural frequencies change during shaking to both damaged and undamaged buildings (see Casey Bradford’s dissertation). Vanessa Heckman, Monica Kohler, and I developed a novel new technique to detect and locate fracture of moment resisting connections in steel buildings. High-frequency waves (> 100 Hz) are radiated throughout a building frame when fracture of brittle weld occurs. Although these welded connections are important to the integrity of a steel building, it is currently very difficult (expensive) to detect when connections fail. Our technique uses seismic records to detect and locate these weld fractures. Obtaining recordings of ground motion have been facilitated by the development of crowd-sourced seismic networks. Traditional seismic networks consists of instruments that are installed and maintained by personnel working for the network operator. In contrast, seismic stations in a crowd-sourced network are operated by others. These others can include volunteers or they may also include professionals at collaborating agencies (e.g., rail transportation agencies, utilities, etc.). The fact that all smart phones have mems accelerometers means that we may one day receive seismic records from millions of cell phones. These records can give us a much more detailed picture of the seismic wavefield as it propagates through California. The most revolutionary aspect of crowd sourced networks is likely to come from building monitoring. Some day in the not-too-distant future there will be a time when the vibrational history of virtually every building will be recorded for significant earthquakes. An additional benefit of the Community Seismic Network is that it will send many real-time estimates of shaking intensity. This will allow for the construction of far more detailed ShakeMaps than is currently feasible. Furthermore, real-time shaking messages will greatly help to make ShakeAlert faster and more reliable. Caltech Civil Engineers (Monica Kohler and Ming Hei Cheng) are collaborating with Caltech Seismologist, Prof. Rob Clayton, and Caltech Computer Scientist, Prof. K. Mani Chandy, to develop the Community Seismic Network (CSN). This exciting project is funded by the Gordon and Betty Moore Foundation. Currently, there are more than 500 3-component accelerometers that continuously telemeter acceleration data. 100 of these are located on the campuses of the Los Angeles Unified School District. |
Students |
Nineke Oerlemans, 1999, MS in Geophysics from Utrecht Univ. (co-advised with H. Paulssen), Sorting Source Parameters to Produce Coherent Record Sections pdf Filippos Filippitzi, Ae, Using the Community Seismic Network to track the health of buildings (co-advising with Monica Kohler) |
| Courses Taught |
ME 35c Statics and Dynamics. 9 units (3-0-6); Prerequisites: Ma 1 abc, Ph 1 abc, Introduction to analysis of stress and strain in engineering. |
Awards and Honors |
Seismological Society of America (President 1993-1995) |
| Publications |
|
| Presentations |
Heaton, T., A. Elbanna, B. Aagaard, and D. Smith, 2013, Implications of Strong-Rate-Weakening Friction for the Length-Scale Dependence of the Strength of the Crust; Why Earthquakes Are so Gentle, IASPEI, Gothenburg, Sweden pdf Heaton, T., A. Olsen, M. Yamada, B. Aagaard, 2013, Statistical Characteristics of Earthquake Ground Motion; Has PBEE Broken the Power Law?, presentation in Kyoto Japan School of Architecture. pdf Heaton, T., Boese, M., Hauksson, E., Allen, G., Cua, G., and M. Yamada, 2012, Earthquake Alerting in California, presented at ETH. pdf Heaton, T., Clayton, R., Chandy, K.M., Kohler, M., Cheng, M.H., Cochrane, E., Lawrence, J., 2012, Community Seismic Network, presented at ETH. pdf Heaton, T., M. Kohler. V. Heckman. M.H. Cheng, C. Bradford, B. Aagaard, J. Clinton, J. Favela, 2012, Using Seismometers to Detect Damage in Buildings. pdf Heaton, T. and Jing Yang, 2009, Seismological Society of America Annual Meeting, Simulated Deformations of Seattle High-Rise Buildings from a Hypothetical Giant Cascadian Earthquake. pdf Heaton, T.H., A. Elbanna, and J. Marsden, 2008 Fall AGU, Size dependence of stress in materials with self-organized critical prestress. pdf Heaton, T., A. Olsen, J. Yang, M. Yamada, 2008 World Conference on Earthquake Engineering, Bejing, Simulations of Flexible Buildings in Large Earthquakes ppt Heaton, T., G. Cua, M. Yamada, M. Böse, 2008, NRC Committee on Seismology and Geodynamics, Creating the Virtual Seismologist for seismic early warning. ppt |