Search USGS⌕ Search

Geology topics

Stephen H. Hartzell

Publications and source records attributed to Stephen H. Hartzell.

At least 37 records · Page 2Linked to original sources
USGS2015

Validation of simulated earthquake ground motions based on evolution of intensity and frequency content

Simulated earthquake ground motions can be used in many recent engineering applications that require time series as input excitations. However, applicability and validation of simulations are subjects of debate in the seismological and engineering communities. We propose a validation methodology at the waveform level and directly based on characteristics that are expected to influence most structural and geotechnical response parameters. In particular, three time-dependent validation metrics are used to evaluate the evolving intensity, frequency, and bandwidth of a waveform. These validation metrics capture nonstationarities in intensity and frequency content of waveforms, making them ideal to address nonlinear response of structural systems. A two-component error vector is proposed to quantify the average and shape differences between these validation metrics for a simulated and recorded ground-motion pair. Because these metrics are directly related to the waveform characteristics, they provide easily interpretable feedback to seismologists for modifying their ground-motion simulation models. To further simplify the use and interpretation of these metrics for engineers, it is shown how six scalar key parameters, including duration, intensity, and predominant frequency, can be extracted from the validation metrics. The proposed validation methodology is a step forward in paving the road for utilization of simulated ground motions in engineering practice and is demonstrated using examples of recorded and simulated ground motions from the 1994 Northridge, California, earthquake.

California↗
USGS2015

Ground motion simulation for the 23 August 2011, Mineral, Virginia earthquake using physics-based and stochastic broadband methods

Three broadband simulation methods are used to generate synthetic ground motions for the 2011 Mineral, Virginia, earthquake and compare with observed motions. The methods include a physics‐based model by Hartzell et al. (1999, 2005), a stochastic source‐based model by Boore (2009), and a stochastic site‐based model by Rezaeian and Der Kiureghian (2010, 2012). The ground‐motion dataset consists of 40 stations within 600 km of the epicenter. Several metrics are used to validate the simulations: (1) overall bias of response spectra and Fourier spectra (from 0.1 to 10 Hz); (2) spatial distribution of residuals for GMRotI50 peak ground acceleration (PGA), peak ground velocity, and pseudospectral acceleration (PSA) at various periods; (3) comparison with ground‐motion prediction equations (GMPEs) for the eastern United States. Our results show that (1) the physics‐based model provides satisfactory overall bias from 0.1 to 10 Hz and produces more realistic synthetic waveforms; (2) the stochastic site‐based model also yields more realistic synthetic waveforms and performs superiorly for frequencies greater than about 1 Hz; (3) the stochastic source‐based model has larger bias at lower frequencies (<0.5&thinsp;&thinsp;Hz) and cannot reproduce the varying frequency content in the time domain. The spatial distribution of GMRotI50 residuals shows that there is no obvious pattern with distance in the simulation bias, but there is some azimuthal variability. The comparison between synthetics and GMPEs shows similar fall‐off with distance for all three models, comparable PGA and PSA amplitudes for the physics‐based and stochastic site‐based models, and systematic lower amplitudes for the stochastic source‐based model at lower frequencies (<0.5&thinsp;&thinsp;Hz).

Bulletin of the Seismological Society of America↗
USGS2015

Ground motion-simulations of 1811-1812 New Madrid earthquakes, central United States

We performed a suite of numerical simulations based on the 1811&ndash;1812 New Madrid seismic zone (NMSZ) earthquakes, which demonstrate the importance of 3D geologic structure and rupture directivity on the ground‐motion response throughout a broad region of the central United States (CUS) for these events. Our simulation set consists of 20 hypothetical earthquakes located along two faults associated with the current seismicity trends in the NMSZ. The hypothetical scenarios range in magnitude from M 7.0 to 7.7 and consider various epicenters, slip distributions, and rupture characterization approaches. The low‐frequency component of our simulations was computed deterministically up to a frequency of 1 Hz using a regional 3D seismic velocity model and was combined with higher‐frequency motions calculated for a 1D medium to generate broadband synthetics (0&ndash;40 Hz in some cases). For strike‐slip earthquakes located on the southwest&ndash;northeast‐striking NMSZ axial arm of seismicity, our simulations show 2&ndash;10 s period energy channeling along the trend of the Reelfoot rift and focusing strong shaking northeast toward Paducah, Kentucky, and Evansville, Indiana, and southwest toward Little Rock, Arkansas. These waveguide effects are further accentuated by rupture directivity such that an event with a western epicenter creates strong amplification toward the northeast, whereas an eastern epicenter creates strong amplification toward the southwest. These effects are not as prevalent for simulations on the reverse‐mechanism Reelfoot fault, and large peak ground velocities (>40&thinsp;&thinsp;cm/s) are typically confined to the near‐source region along the up‐dip projection of the fault. Nonetheless, these basin response and rupture directivity effects have a significant impact on the pattern and level of the estimated intensities, which leads to additional uncertainty not previously considered in magnitude estimates of the 1811&ndash;1812 sequence based only on historical reports. The region covered by our simulation domain encompasses a large portion of the CUS centered on the NMSZ, including several major metropolitan areas. Based on our simulations, more than eight million people living and working near the NMSZ would experience potentially damaging ground motion and modified Mercalli intensities ranging from VI to VIII if a repeat of the 1811&ndash;1812 earthquakes occurred today. Moreover, the duration of strong ground shaking in the greater Memphis metropolitan area could last from 30 to more than 60 s, depending on the magnitude and epicenter.

Arkansas, Illinois, Kentucky, Missouri, Tennessee↗
USGS2015

The use of wavenumber normalization in computing spatially averaged coherencies (KRSPAC) of microtremor data from asymmetric arrays

The SPAC method of processing microtremor noise observations for estimation of Vs profiles has a limitation that the array has circular or triangular symmetry in order to allow spatial (azimuthal) averaging of inter-station coherencies over a constant station separation. Common processing methods allow for station separations to vary by typically &plusmn;10% in the azimuthal averaging before degradation of the SPAC spectrum is excessive. A limitation on use of high-wavenumbers in inversions of SPAC spectra to Vs profiles has been the requirement for exact array symmetry to avoid loss of information in the azimuthal averaging step. In this paper we develop a new wavenumber-normalised SPAC method (KRSPAC) where instead of performing averaging of sets of coherency versus frequency spectra and then fitting to a model SPAC spectrum, we interpolate each spectrum to coherency versus k.r, where k and r are wavenumber and station separation respectively, and r may be different for each pair of stations. For fundamental mode Rayleigh-wave energy the model SPAC spectrum to be fitted reduces to Jo(kr). The normalization process changes with each iteration since k is a function of frequency and phase velocity and hence is updated each iteration. The method proves robust and is demonstrated on data acquired in the Santa Clara Valley, CA, (Site STGA) where an asymmetric array having station separations varying by a factor of 2 is compared with a conventional triangular array; a 300-mdeep borehole with a downhole Vs log provides nearby ground truth. The method is also demonstrated on data from the Pleasanton array, CA, where station spacings are irregular and vary from 400 to 1200 m. The KRSPAC method allows inversion of data using kr (unitless) values routinely up to 30, and occasionally up to 60. Thus despite the large and irregular station spacings, this array permits resolution of Vs as fine as 15 m for the near-surface sediments, and down to a maximum depth of 2.5 km.

Conference Paper↗
USGS2014

Relation of landslides triggered by the Kiholo Bay earthquake to modeled ground motion

The 2006 Kiholo Bay, Hawaii, earthquake triggered high concentrations of rock falls and slides in the steep canyons of the Kohala Mountains along the north coast of Hawaii. Within these mountains and canyons a complex distribution of landslides was triggered by the earthquake shaking. In parts of the area, landslides were preferentially located on east‐facing slopes, whereas in other parts of the canyons no systematic pattern prevailed with respect to slope aspect or vertical position on the slopes. The geology within the canyons is homogeneous, so we hypothesize that the variable landslide distribution is the result of localized variation in ground shaking; therefore, we used a state‐of‐the‐art, high‐resolution ground‐motion simulation model to see if it could reproduce the landslide‐distribution patterns. We used a 3D finite‐element analysis to model earthquake shaking using a 10 m digital elevation model and slip on a finite‐fault model constructed from teleseismic records of the mainshock. Ground velocity time histories were calculated up to a frequency of 5 Hz. Dynamic shear strain also was calculated and compared with the landslide distribution. Results were mixed for the velocity simulations, with some areas showing correlation of landslide locations with peak modeled ground motions but many other areas showing no such correlation. Results were much improved for the comparison with dynamic shear strain. This suggests that (1) rock falls and slides are possibly triggered by higher frequency ground motions (velocities) than those in our simulations, (2) the ground‐motion velocity model needs more refinement, or (3) dynamic shear strain may be a more fundamental measurement of the decoupling process of slope materials during seismic shaking.

Hawai'i↗
USGS2014

Finite-fault slip model of the 2011 M w 5.6 Prague, Oklahoma earthquake from regional waveforms

The slip model for the 2011 M w 5.6 Prague, Oklahoma, earthquake is inferred using a linear least squares methodology. Waveforms of six aftershocks recorded at 21 regional stations are used as empirical Green's functions (EGFs). The solution indicates two large slip patches: one located around the hypocenter with a depth range of 3–5.5 km; the other located to the southwest of the epicenter with a depth range from 7.5 to 9.5 km. The total moment of the solution is estimated at 3.37 × 10 24 dyne cm (M w 5.65). The peak slip and average stress drop for the source at the hypocenter are 70 cm and 90 bars, respectively, approximately one half the values for the M w 5.8 2011 Mineral, Virginia, earthquake. The stress drop averaged over all areas of slip is 16 bars. The relatively low peak slip and stress drop may indicate an induced component in the origin of the Prague earthquake from deep fluid injection.

Oklahoma↗
USGS2013

Ground motion in the presence of complex topography: Earthquake and ambient noise sources

To study the influence of topography on ground motion, eight seismic recorders were deployed for a period of one year over Poverty Ridge on the east side of the San Francisco Bay Area, California. This location is desirable because of its proximity to local earthquake sources and the significant topographic relief of the array (439 m). Topographic amplification is evaluated as a function of frequency using a variety of methods, including reference‐site‐based spectral ratios and single‐station horizontal‐to‐vertical spectral ratios using both shear waves from earthquakes and ambient noise. Field observations are compared with the predicted ground motion from an accurate digital model of the topography and a 3D local velocity model. Amplification factors from the theoretical calculations are consistent with observations. The fundamental resonance of the ridge is prominently observed in the spectra of data and synthetics; however, higher‐frequency peaks are also seen primarily for sources in line with the major axis of the ridge, perhaps indicating higher resonant modes. Excitations of lateral ribs off of the main ridge are also seen at frequencies consistent with their dimensions. The favored directions of resonance are shown to be transverse to the major axes of the topographic features.

California↗
USGS2013

Variability in wood-frame building damage using broad-band synthetic ground motions: a comparative numerical study with recorded motions

Earthquake damage to light-frame wood buildings is a major concern for North America because of the volume of this construction type. In order to estimate wood building damage using synthetic ground motions, we need to verify the ability of synthetically generated ground motions to simulate realistic damage for this structure type. Through a calibrated damage potential indicator, four different synthetic ground motion models are compared with the historically recorded ground motions at corresponding sites. We conclude that damage for sites farther from the fault (>20 km) is under-predicted on average and damage at closer sites is sometimes over-predicted.

Journal of Earthquake Engineering↗
USGS2013

Rupture model of the 2011 Mineral, Virginia, earthquake from teleseismic and regional waveforms

We independently invert teleseismic P waveforms and regional crustal phases to examine the finite fault slip model for the 2011 M w 5.8 Mineral, Virginia, earthquake. Theoretical and empirical Green's functions are used for the teleseismic and regional models, respectively. Both solutions show two distinct sources each about 2&thinsp;km across and separated by 2.5&thinsp;km. The source at the hypocenter is more localized in the regional model leading to a higher peak slip of 130&thinsp;cm and higher average stress drop of 250 bars compared with 86&thinsp;cm and 150 bars for the same source in the teleseismic model. Both sources are centered at approximately 8&thinsp;km depth in the regional model, largely below the aftershock distribution. In the teleseismic model, the sources extend updip to approximately 6&thinsp;km depth, into the depth range of the aftershocks. The rupture velocity is not well resolved but appears to be near 2.7&thinsp;km/s.

Geophysical Research Letters↗
USGS2013

Rupture history of the 2008 M w 7.9 Wenchuan, China, earthquake: Evaluation of separate and joint inversions of geodetic, teleseismic, and strong-motion data

An extensive data set of teleseismic and strong-motion waveforms and geodetic offsets is used to study the rupture history of the 2008 Wenchuan, China, earthquake. A linear multiple-time-window approach is used to parameterize the rupture. Because of the complexity of the Wenchuan faulting, three separate planes are used to represent the rupturing surfaces. This earthquake clearly demonstrates the strengths and limitations of geodetic, teleseismic, and strong-motion data sets. Geodetic data (static offsets) are valuable for determining the distribution of shallower slip but are insensitive to deeper faulting and reveal nothing about the timing of slip. Teleseismic data in the distance range 30&deg;&ndash;90&deg; generally involve no modeling difficulties because of simple ray paths and can distinguish shallow from deep slip. Teleseismic data, however, cannot distinguish between different slip scenarios when multiple fault planes are involved because steep takeoff angles lead to ambiguity in timing. Local strong-motion data, on the other hand, are ideal for determining the direction of rupture from directivity but can easily be over modeled with inaccurate Green&rsquo;s functions, leading to misinterpretation of the slip distribution. We show that all three data sets are required to give an accurate description of the Wenchuan rupture. The moment is estimated to be approximately 1.0 &times; 10 21 N &middot; m with the slip characterized by multiple large patches with slips up to 10 m. Rupture initiates on the southern end of the Pengguan fault and proceeds unilaterally to the northeast. Upon reaching the cross-cutting Xiaoyudong fault, rupture of the adjacent Beichuan fault starts at this juncture and proceeds bilaterally to the northeast and southwest.

Wenchuan↗
USGS2012

Seismic velocity model of the central United States (Version 1): Description and simulation of the 18 April 2008 Mt. Carmel, Illinois, Earthquake

We have developed a new three‐dimensional seismic velocity model of the central United States (CUSVM) that includes the New Madrid Seismic Zone (NMSZ) and covers parts of Arkansas, Mississippi, Alabama, Illinois, Missouri, Kentucky, and Tennessee. The model represents a compilation of decades of crustal research consisting of seismic, aeromagnetic, and gravity profiles; geologic mapping; geophysical and geological borehole logs; and inversions of the regional seismic properties. The density, P ‐ and S ‐wave velocities are synthesized in a stand‐alone spatial database that can be queried to generate the required input for numerical seismic‐wave propagation simulations. We test and calibrate the CUSVM by simulating ground motions of the 18 April 2008 M w 5.4 Mt. Carmel, Illinois, earthquake and comparing the results with observed records within the model area. The selected stations in the comparisons reflect different geological site conditions and cover distances ranging from 10 to 430 km from the epicenter. The results, based on a qualitative and quantitative goodness‐of‐fit (GOF) characterization, indicate that both within and outside the Mississippi Embayment the CUSVM reasonably reproduces: (1) the body and surface‐wave arrival times and (2) the observed regional variations in ground‐motion amplitude, cumulative energy, duration, and frequency content up to a frequency of 1.0 Hz. In addition, we discuss the probable structural causes for the ground‐motion patterns in the central United States that we observed in the recorded motions of the 18 April Mt. Carmel earthquake.

Illinois↗
USGS2012

Importance of model parameterization in finite fault inversions: Application to the 1974 Mw 8.0 Peru Earthquake

The spatial and temporal slip distributions for the October 3, 1974 ( M w = 8.0), Peru subduction zone earthquake and its largest aftershock on November 9 ( M s = 7.1) are calculated and analyzed in terms of the inversion parameterization and tectonic significance. Teleseismic, long-period World-Wide Standard Seismograph Network, P and SH waveforms are inverted to obtain the rupture histories. We demonstrate that erroneous results are obtained if a parameterization is used that does not allow for a sufficiently complex source, involving spatial variation in slip amplitude, risetime, and rupture time. The inversion method utilizes a parameterization of the fault that allows for a discretized source risetime and rupture time. Well-located aftershocks recorded on a local network have the same general pattern as teleseismically determined hypocenters and help to constrain the geometry of the subduction zone. For the main shock a hinged fault is preferred having a shallow plane with a dip of 11° and a deeper, landward plane with a dip of 30°. The preferred nucleation depth lies between 11 and 15 km. A bilateral rupture is obtained with two major concentrations of slip, one 60 to 70 km to the northwest of the epicenter and a second 80 to 100 km to the south and southeast of the epicenter. For these source regions, risetimes vary from 6 to 18 s. Our estimates of risetimes are consistent with the time for the rupture to traverse the dominant local asperity. The slip distribution for the November 9 aftershock falls within a conspicuous hole in the main shock rupture pattern, near the hypocenter of the main shock. The November 9 event has a simple risetime function with a duration of 2 s. Aftershocks recorded by the local network are shown to cluster near the hypocenter of the impending November 9 event and downdip from the largest main shock source region. Slip during the main shock is concentrated at shallow depths above 15 km and extends updip from the hypocenter to near the plate boundary at the trench axis. The large amount of slip at shallow depths is attributed to the absence of any significant accretionary wedge of sediments, and the relatively young age and high convergence rate of the subducted plate, which results in good seismic coupling near the trench axis.

Journal of Geophysical Research B: Solid Earth↗
USGS2011

Model and parametric uncertainty in source-based kinematic models of earthquake ground motion

Four independent ground-motion simulation codes are used to model the strong ground motion for three earthquakes: 1994 M w 6.7 Northridge, 1989 M w 6.9 Loma Prieta, and 1999 M w 7.5 Izmit. These 12 sets of synthetics are used to make estimates of the variability in ground-motion predictions. In addition, ground-motion predictions over a grid of sites are used to estimate parametric uncertainty for changes in rupture velocity. We find that the combined model uncertainty and random variability of the simulations is in the same range as the variability of regional empirical ground-motion data sets. The majority of the standard deviations lie between 0.5 and 0.7 natural-log units for response spectra and 0.5 and 0.8 for Fourier spectra. The estimate of model epistemic uncertainty, based on the different model predictions, lies between 0.2 and 0.4, which is about one-half of the estimates for the standard deviation of the combined model uncertainty and random variability. Parametric uncertainty, based on variation of just the average rupture velocity, is shown to be consistent in amplitude with previous estimates, showing percentage changes in ground motion from 50% to 300% when rupture velocity changes from 2.5 to 2.9 km/s. In addition, there is some evidence that mean biases can be reduced by averaging ground-motion estimates from different methods.

Bulletin of the Seismological Society of America↗
USGS2009

Data files for ground-motion simulations of the 1906 San Francisco earthquake and scenario earthquakes on the Northern San Andreas Fault

This data set contains results from ground-motion simulations of the 1906 San Francisco earthquake, seven hypothetical earthquakes on the northern San Andreas Fault, and the 1989 Loma Prieta earthquake. The bulk of the data consists of synthetic velocity time-histories. Peak ground velocity on a 1/60th degree grid and geodetic displacements from the simulations are also included. Details of the ground-motion simulations and analysis of the results are discussed in Aagaard and others (2008a,b).

Data Series↗
USGS2008

Simulated ground motion in Santa Clara Valley, California, and vicinity from M≥6.7 scenario earthquakes

Models of the Santa Clara Valley (SCV) 3D velocity structure and 3D finite-difference software are used to predict ground motions from scenario earthquakes on the San Andreas (SAF), Monte Vista/Shannon, South Hayward, and Calaveras faults. Twenty different scenario ruptures are considered that explore different source models with alternative hypocenters, fault dimensions, and rupture velocities and three different velocity models. Ground motion from the full wave field up to 1 Hz is exhibited as maps of peak horizontal velocity and pseudospectral acceleration at periods of 1, 3, and 5 sec. Basin edge effects and amplification in sedimentary basins of the SCV are observed that exhibit effects from shallow sediments with relatively low shear-wave velocity (330 m/sec). Scenario earthquakes have been simulated for events with the following magnitudes: (1) M 6.8&ndash;7.4 Calaveras sources, (2) M 6.7&ndash;6.9 South Hayward sources, (3) M 6.7 Monte Vista/Shannon sources, and (4) M 7.1&ndash;7.2 Peninsula segment of the SAF sources. Ground motions are strongly influenced by source parameters such as rupture velocity, rise time, maximum depth of rupture, hypocenter, and source directivity. Cenozoic basins also exert a strong influence on ground motion. For example, the Evergreen Basin on the northeastern side of the SCV is especially responsive to 3&ndash;5-sec energy from most scenario earthquakes. The Cupertino Basin on the southwestern edge of the SCV tends to be highly excited by many Peninsula and Monte Vista fault scenarios. Sites over the interior of the Evergreen Basin can have long-duration coda that reflect the trapping of seismic energy within this basin. Plausible scenarios produce predominantly 5-sec wave trains with greater than 30 cm/sec sustained ground-motion amplitude with greater than 30 sec duration within the Evergreen Basin.

California↗
USGS1996

Site-response maps for the Los Angeles region based on earthquake ground motions

Ground-motion records from aftershocks of the 1994 Northridge earthquake and main-shock records from the 1971 San Fernando, 1987 Whittier Narrows, 1991 Sierra Madre, and 1994 Northridge earthquakes are used to estimate site response in the urban Los Angeles, California, area. Two frequency bands are considered, 0.5-1.5 Hz and 2.0-6.0 Hz. Instrument characteristics prevented going to lower frequencies, and frequencies above 6.0 Hz are less important to the building inventory. Site response determined at the instrumented locations is associated with the surficial geology and contoured to produce a continuous spatial estimation of site response. The maps in this report are preliminary and will evolve as more data become available and more analysis is done.

California↗