Earthquake Science Center Seminars

Ross will use QuakeCaster and Fishnet Stocking Stress, demonstration tools he developed with students, to probe the fundamental riddle of earthquakes: why quakes are sporadic when their driving forces are steady. The demo also shows how earthquakes converse by the transfer of static and dynamic stress, an insight that helps crack the difficult problem of earthquake forecasting. Ross will then delve into the strange world of earthquake triggering and inhibition. He will argue that while most earthquakes interact over distances of a few fault rupture dimensions, rarely, megaquakes can engulf the entire globe in large aftershocks for days to a week.

The November 14th, 2016, Mw 7.8 Kaikoura earthquake generated significant shaking and damage throughout a large part of central New Zealand. The recovery effort has been distributed over a very large area and the dissemination of information about the potential for future shaking has been an integral part of the GNS Science response. Through the nature of the event and the varied recovery efforts we have been required to provide model results that spanned from aftershock probability tables through to detailed and specific engineering information. Additionally, the main shock triggered three slow slip events (SSE) on the Hikurangi subduction zone that were unique in character in our approximately 20 years of observations. These SSE have provided a difficult and interesting challenge to the ongoing forecasting efforts.

Our overriding philosophy remains that no single model sufficiently captures our current understanding of earthquake occurrence and that by using a hybrid of multiple models, we best capture the uncertainty in our knowledge of future quakes. To this point, and similar to past New Zealand earthquakes, we have used a hybrid forecast model based on the the STEP, ETAS, and EEPAS models, which we have combined with a range of long-term models, including a new strain-rate based model. We have used this hybrid model to produce forecasts from 1-day to 100-years. New Zealand end-users have become increasingly sophisticated in their uptake of forecasting information with increasing requests and needs from across sectors. A noticeable change in the Kaikoura response was the desire for forecast information not only to provided in seismic hazard terms, but in very specific risk terms for government and industry decision making.

In the past seven years, New Zealand has experienced around 15 large earthquakes which have required a forecasting response. In this presentation, I will discuss our forecast models, the engineering outputs developed, our efforts to estimate the impact of the SSE on the potential for triggered earthquakes.

The substantial increase in earthquake rate observed in Oklahoma during the last decade has largely been attributed to the disposal of wastewater from energy production activities. While existing sparse regional networks are able to capture the occurrence of most M3 earthquakes and some smaller magnitude events, micro to minor earthquakes typically cannot be detected. To enable the detection of microseismicity, we deployed a temporary array of >1,800 nodal seismometers in northern Oklahoma. The LArge-n Seismic Survey in Oklahoma (LASSO) array operated for approximately one month, covering a 25-km-by-32-km region with a nominal station spacing of ~400 m. We develop a local earthquake catalog from data recorded by the LASSO array, which allows us to investigate sequences of induced seismicity in this region of active wastewater injection with unprecedented clarity. A two-stage earthquake detection method identifies more than 13,000 earthquakes recorded by the array, a factor of ~140 increase over the regional earthquake catalog. We explore frequency-magnitude relationships and the spatiotemporal evolution of seismicity in our catalog. The locations and orientations of select active faults, as inferred from seismicity, will also be presented.

The Hayward Fault presents significant seismic hazard to the eastern San Francisco Bay Area (the "East Bay"). We report on a high-performance computing (HPC) simulation of strong ground motions for a magnitude (M) 7.0 scenario earthquake on the Hayward Fault resolved to 5 Hz and higher. We computed seismic motions obeying physics-based wave propagation in a regional-scale, three-dimensional (3D) Earth model with surface topography with the SW4 finite difference code. Both plane-layered (one-dimensional, 1D) and 3D earth models were considered, with 3D sub-surface material properties and topography sub-sampled from a 3D model by the USGS. The resulting ground motion intensities (peak ground acceleration and velocity, RotD50 spectral accelerations) are in good agreement with ground motion models (GMM) based on empirical data. Median intensities for sites across the domain are within the reported uncertainties of the Abrahamson et al. (2014) GMM across spectral periods 0.1-10 seconds (frequencies 0.1-10 Hz). The variance of ground motion intensity measurements (GMIM's) residuals ranges 0.4-0.6 natural log units, representing the site-to-site, within event variability for this single rupture. These results demonstrate that earthquake simulations with fully 3D deterministic wave propagation on world-class HPC platforms provides ground motions consistent with empirical models for frequencies of relevance to engineering analysis.

There is a building boom for tall buildings for West Coast Cities; daring architectural designs trumpet that they are designed to withstand the 2,500-yr earthquake shaking. In this talk, I explore whether or not these claims are scientifically based, or are we being used as "useful idiots" to facilitate the ambitions of developers? Cutting through the claims of current high-rise development is surprisingly difficult. Technical reports describing the attributes of real buildings are mostly proprietary and the deliberations of peer-review committees are secret. To help better understand the collapse resistance of typical tall buildings I have worked with my colleagues and students to simulate the response of steel moment-resisting-frame buildings designed to meet building codes that have evolved considerably since the 1950's. We have focused on the problem of building collapse; this occurs when steel columns become tilted enough that gravitational forces cause bending moments (as opposed to axial forces for vertical columns) that exceed the bending yield capacity of steel sections. I will show that the combination of high peak ground velocity (pgv) and high peak displacement (pgd) are much better predictors of collapse than is 5% spectral acceleration, which is currently the focus of the National Probabilistic Seismic Hazard maps (NPSHA). High pgv (>75 cm/s) induces plastic yielding and large pgd (>75 cm) causes column tilts that are gravitationally unstable. I will further show that 70% damped response spectra are another alternate collapse predictor. I also discuss the problem of statistical analysis of pgv, pgd. I argue that pgd is poorly characterized by earthquake magnitude and that its occurrence is described by a heavy-tailed power law with an undefined variance. I argue that current statements that tall buildings are constructed for 2,500-yr shaking are misleading and are designed to lull the public into believing that tall buildings are safe in future large earthquakes. I will show how the evolution of building codes over the past 60 years has affected the intended collapse resistance of tall buildings (surprisingly little). I will also show that pre-Northridge steel buildings are much less collapse resistant due to mistakes in the design of critical beam/column welded connections.

The interaction between the seismic portion of faults and their deeper roots is central to understanding the mechanics of earthquake cycles. It is well established that faults are highly localized within the cold, brittle upper crust, but less is known about fault structure in the warmer, more ductile, lower crust and upper mantle. Increasing temperature with depth causes two transitions in behavior: a frictional transition from seismic to aseismic fault behavior, and a transition from brittle to ductile bulk deformation. To explore the interaction between these two transitions, we simulate earthquake cycles on a 2D strike-slip fault, representative of an intercontinental fault like the San Andreas. All phases of the earthquake cycle are simulated, allowing the model to spontaneously generate earthquakes, and to capture frictional afterslip and postseismic and interseismic viscous flow. The fault is represented with rate-and-state friction. Off-fault viscoelasticity is represented with a power-law rheology for dislocation creep, using experimentally-based flow laws for feldspar in the crust and olivine in the upper mantle. The power-law rheology involves an effective viscosity that is a function of temperature and stress, and therefore varies both spatially and temporally. Since the behavior of the system is highly sensitive to temperature, we explore a range of ambient geotherms and additionally incorporate heat generation through frictional and viscous shear heating.

In all models, a zone of lower effective viscosity forms near the fault, which is narrow in the lower crust and which broadens with depth, and most bulk viscous flow occurs within this zone. Large earthquakes, such as the Hector Mine earthquake, have sometimes been observed to cause an increase in effective viscosity. In our model, we only see this type of behavior in simulations with the warmest ambient geotherm considered. Additionally, we find that shear heating significantly decreases the depth of the brittle-ductile transition, and that frictional and viscous shear heating contribute roughly equally. Finally, we explore differences in surface deformation and surface heat flux produced by these models.

Observational seismology is an increasingly data-rich field in which machine learning techniques show significant potential. In this seminar I focus on two specific examples where simple algorithms and concepts from machine learning prove useful in characterizing earthquake source properties and hazard. I will begin by describing a new framework for empirical Ground Motion Prediction Equations (GMPEs) based on a supervised learning algorithm known as a Random Forest. To demonstrate its utility in earthquake hazard analysis, I use the Random Forest GMPE to measure the influence of dynamic stress drop on the observed peak ground accelerations of moderate magnitude earthquakes in the San Francisco Bay Area. In the second part of this seminar, I will discuss ongoing collaborative efforts to systematically analyze the P-waveform features of large magnitude earthquakes for potential signatures of nucleation and rupture onset characteristics that correlate with event size. Many real-time earthquake early warning algorithms implicitly assume a form of rupture determinism in which the earthquake magnitude can be rapidly estimated from a small snapshot of waveform data, yet it remains unclear how viable these approaches are for large-magnitude events. Through analyses of a large dataset of Japanese Kik-Net borehole waveforms, we hope to gain insight into the physics of earthquake rupture and to provide unbiased observational constraints for real-time hazard applications.

To date, reverse time migration (RTM) is one of the most accurate imaging tools available in the energy industry to deliver accurate subsurface reflectivity distributions. In comparison to other imaging tools, it allows us to take complicated wave physics into account, and map complex Earth's structures in great details. In this talk, I will present our strategies towards promoting RTM from acoustic to elastic regime. The key challenges to be addressed are: (1) how to accurately decompose elastic P and S wavefields; and (2) what imaging conditions can be used to deal with multi-component elastic wavefields. Over the past several years, we have developed algorithms to tackle these challenges, which have been tested on 2-D and 3-D synthetic experiments. In the first part of this talk, I will present the applications of elastic RTM to improve our capability for delineating reflectivity distributions in complex subsurface environment. In the second part, I will present its applications to imaging earthquake rupture, fracture propagation as well as monitoring microseismicity. In these cases, we do not need to invoke assumptions, such as 1-D velocity models, single rectangular fault planes, and P/S arrival picking. Synthetic results demonstrate that the new technology allows us to characterize complicated temporal and spatial evolution of earthquake rupture processes.