Geology topics

Gregory C. Beroza

Publications and source records attributed to Gregory C. Beroza.

18 recordsLinked to original sources

USGS2025

Divide and conquer: Separating the two probabilities in seismic phase picking

There are two fundamental probabilities in the seismic phase picking process – the probability of the existence of a seismic phase (detection probability) and the probability of correctly identifying the phase arrival time (timing probability). The nearly ubiquitous approach in developing deep learning phase picking models is to use a kernel, such as a truncated Gaussian, to mask the labeled phase arrival time, and train a segmentation model. Once a model is trained, the times of the peaks in the output are taken as phase arrival times (picks) and the height of the peaks are taken as “probability” of the picks. Here, we show that this “probability” represents neither the detection nor the timing probabilty because this approach forces the output to follow the shape of the kernel. We introduce an approach using two models to estimate these two distinct probabilities. We use a binary classifier with a calibrated confidence to address the detection probability and a multi-class classifier to obtain a probability mass function to address the timing probability. This new approach makes the deep learning-based phase picking process more interpretable and gives us options to logically control seismic monitoring workflows.

Geophysical Journal International↗

USGS2025

Overview of The SCEC/USGS Community Stress Drop Validation Study using the 2019 Ridgecrest earthquake sequence

We present initial findings from the ongoing Community Stress Drop Validation Study to compare spectral stress‐drop estimates for earthquakes in the 2019 Ridgecrest, California, sequence. This study uses a unified dataset to independently estimate earthquake source parameters through various methods. Stress drop, which denotes the change in average shear stress along a fault during earthquake rupture, is a critical parameter in earthquake science, impacting ground motion, rupture simulation, and source physics. Spectral stress drop is commonly derived by fitting the amplitude‐spectrum shape, but estimates can vary substantially across studies for individual earthquakes. Sponsored jointly by the U.S. Geological Survey and the Statewide (previously, Southern) California Earthquake Center our community study aims to elucidate sources of variability and uncertainty in earthquake spectral stress‐drop estimates through quantitative comparison of submitted results from independent analyses. The dataset includes nearly 13,000 earthquakes ranging from M 1 to 7 during a two‐week period of the 2019 Ridgecrest sequence, recorded within a 1° radius. In this article, we report on 56 unique submissions received from 20 different groups, detailing spectral corner frequencies (or source durations), moment magnitudes, and estimated spectral stress drops. Methods employed encompass spectral ratio analysis, spectral decomposition and inversion, finite‐fault modeling, ground‐motion‐based approaches, and combined methods. Initial analysis reveals significant scatter across submitted spectral stress drops spanning over six orders of magnitude. However, we can identify between‐method trends and offsets within the data to mitigate this variability. Averaging submissions for a prioritized subset of 56 events shows reduced variability of spectral stress drop, indicating overall consistency in recovered spectral stress‐drop values.

California↗

USGS2022

On the use of high-resolution and deep-learning seismic catalogs for short-term earthquake forecasts: Potential benefits and current limitations

Enhanced earthquake catalogs provide detailed images of evolving seismic sequences. Currently, these data sets take some time to be released but will soon become available in real time. Here, we explore whether and how enhanced seismic catalogs feeding into established short-term earthquake forecasting protocols may result in higher predictive skill. We consider three enhanced catalogs for the 2016–2017 Central Italy sequence, featuring a bulk completeness lower by at least two magnitude units compared to the real-time catalog and an improved hypocentral resolution. We use them to inform a set of physical Coulomb Rate-and-State (CRS) and statistical Epidemic-Type Aftershock Sequence (ETAS) models to forecast the space-time occurrence of M3+ events during the first 6 months of the sequence. We track model performance using standard likelihood-based metrics and compare their skill against the best-performing CRS and ETAS models among those developed with the real-time catalog. We find that while the incorporation of the triggering contributions from new small magnitude detections of the enhanced catalogs is beneficial for both types of forecasts, these models do not significantly outperform their respective near real-time benchmarks. To explore the reasons behind this result, we perform targeted sensitivity tests that show how (a) the typical spatial discretizations of forecast experiments ( 2 km) hamper the ability of models to capture highly localized secondary triggering patterns and (b) differences in earthquake parameters (i.e., magnitude and hypocenters) reported in different catalogs can affect forecast evaluation. These findings will contribute toward improving forecast model design and evaluation strategies for next-generation seismic catalogs.

Journal of Geophysical Research--Solid Earth↗

USGS2022

DevelNet: Earthquake detection on Develocorder films with deep learning: Application to the rangely earthquake control experiment

There exists over a century of instrumental seismic data; however, most seismograms recorded before the 1980s are only available in analog form. Although analog seismograms are of great value, they are underutilized due to the difficulties of making quantitative measurements on the original media and in converting them to digital time series. In this study, we present an alternative workflow, based on deep learning, to reconstruct an earthquake catalog from images of analog data without conversion to vector time series. We trained a convolutional neural network—DevelNet, using synthetic analog data to detect earthquakes on scanned multichannel Develocorder film images. We then developed an image‐based processing workflow to measure arrival times, locate, and determine the magnitudes of earthquakes in the data. We demonstrate the performance of this approach on two years of continuous Develocorder film recordings from the Rangely earthquake control experiment in the mid‐1970s. Our approach detects twice the number of events reported in the original catalog ( Raleigh et al. , 1976 ). This demonstrates that DevelNet efficiently detects earthquakes from Develocorder film scans, performs consistently over time, and is robust to changes in network geometry. Our locations generally agree with the original study, although the automatically measured arrival times are less precise than manual reading, leading to increased location scatter. Our automatic workflow of Develocorder films rivals the performance of skilled analysts in earthquake detection, but with minimal human intervention. This image‐based processing offers a new approach for effectively and efficiently extracting earthquake information from analog seismic data.

Bulletin of the Seismological Society of America↗

USGS2021

Quantifying nuisance ground motion thresholds for induced earthquakes

Hazards from induced earthquakes are a growing concern with a need for effective management. One aspect of that concern is the “nuisance” from unexpected ground motions, which have the potential to cause public alarm and discontent. In this article, we borrow earthquake engineering concepts to quantify the chance of building damage states and adapt them to quantify felt thresholds for induced earthquakes in the Central and Eastern United States. We compare binary data of felt or not-felt reports from the “Did You Feel It” database with ShakeMap ground motion intensity measures (IM) for ∼360 earthquakes. We use a Monte Carlo logistic regression to discern the likelihood of perceiving various degrees of felt intensity, given a particular IM. These best-fit nuisance functions are reported in this article and are readily transferable. Of the shaking types considered, we find that peak ground velocity tends to be the best predictor of a felt earthquake. We also find that felt thresholds tended to decrease with increasing earthquake magnitude, after M ∼3.9. We interpret this effect as related to the duration of the event, where events smaller than M 3.9 are perceived as “impulsive” to the human senses. Improved quantification of the nuisance from induced earthquake ground motions could be utilized in management of the public perception of their causal operations. Although aimed at anthropogenic earthquakes, thresholds we derive could be useful in other realms, such as establishing best practices and protocols for earthquake early warning.

Earthquake Spectra↗

USGS2018

Seismology with dark data: Image-based processing of analog records using machine learning for the rangely earthquake control experiment

Before the digital era, seismograms were recorded in analog form and read manually by analysts. The digital era represents only about 25% of the total time span of instrumental seismology. Analog data provide important constraints on earthquake processes over the long term, and in some cases are the only data available. The media on which analog data are recorded degrades with time and there is an urgent need for cost‐effective approaches to preserve the information they contain. In this study, we work directly with images by constructing a set of image‐based methods for earthquake processing, rather than pursue the usual approach of converting analog data to vector time series. We demonstrate this approach on one month of continuous Develocorder films from the Rangely earthquake control experiment run by the U.S. Geological Survey (USGS). We scan the films into images and compress these into low‐dimensional feature vectors as input to a classifier that separates earthquakes from noise in a defined feature space. We feed the detected event images into a short‐term average/long‐term average (STA/LTA) picker, a grid‐search associator, and a 2D image correlator to measure both absolute arrival times and relative arrival‐time differences between events. We use these measurements to locate the earthquakes using hypoDD. In the month that we studied, we identified 40 events clustered near the injection wells. In the original study, Raleigh et al. (1976) identified only 32 events during the same period. Scanning without vectorizing analog seismograms represents an attractive approach to archiving these perishable data. We demonstrated that it is possible to carry out precision seismology directly on such images. Our approach has the potential for wide application to analog seismograms.

Seismological Research Letters↗

USGS2018

Variabilities in probabilistic seismic hazard maps for natural and induced seismicity in the central and eastern United States

Probabilistic seismic hazard analysis (PSHA) characterizes ground-motion hazard from earthquakes. Typically, the time horizon of a PSHA forecast is long, but in response to induced seismicity related to hydrocarbon development, the USGS developed one-year PSHA models. In this paper, we present a display of the variability in USGS hazard curves due to epistemic uncertainty in its informed submodel using a simple bootstrapping approach. We find that variability is highest in low-seismicity areas. On the other hand, areas of high seismic hazard, such as the New Madrid seismic zone or Oklahoma, exhibit relatively lower variability simply because of more available data and a better understanding of the seismicity. Comparing areas of high hazard, New Madrid, which has a history of large naturally occurring earthquakes, has lower forecast variability than Oklahoma, where the hazard is driven mainly by suspected induced earthquakes since 2009. Overall, the mean hazard obtained from bootstrapping is close to the published model, and variability increased in the 2017 one-year model relative to the 2016 model. Comparing the relative variations caused by individual logic-tree branches, we find that the highest hazard variation (as measured by the 95% confidence interval of bootstrapping samples) in the final model is associated with different ground-motion models and maximum magnitudes used in the logic tree, while the variability due to the smoothing distance is minimal. It should be pointed out that this study is not looking at the uncertainty in the hazard in general, but only as it is represented in the USGS one-year models.

The Leading Edge↗

USGS2016

Constraints on the source parameters of low-frequency earthquakes on the San Andreas Fault

Low-frequency earthquakes (LFEs) are small repeating earthquakes that occur in conjunction with deep slow slip. Like typical earthquakes, LFEs are thought to represent shear slip on crustal faults, but when compared to earthquakes of the same magnitude, LFEs are depleted in high-frequency content and have lower corner frequencies, implying longer duration. Here we exploit this difference to estimate the duration of LFEs on the deep San Andreas Fault (SAF). We find that the M ~ 1 LFEs have typical durations of ~0.2 s. Using the annual slip rate of the deep SAF and the average number of LFEs per year, we estimate average LFE slip rates of ~0.24 mm/s. When combined with the LFE magnitude, this number implies a stress drop of ~10 4 Pa, 2 to 3 orders of magnitude lower than ordinary earthquakes, and a rupture velocity of 0.7 km/s, 20% of the shear wave speed. Typical earthquakes are thought to have rupture velocities of ~80–90% of the shear wave speed. Together, the slow rupture velocity, low stress drops, and slow slip velocity explain why LFEs are depleted in high-frequency content relative to ordinary earthquakes and suggest that LFE sources represent areas capable of relatively higher slip speed in deep fault zones. Additionally, changes in rheology may not be required to explain both LFEs and slow slip; the same process that governs the slip speed during slow earthquakes may also limit the rupture velocity of LFEs.

California↗

USGS2014

Validation of the SCEC broadband platform V14.3 simulation methods using pseudo spectral acceleration data

This paper summarizes the evaluation of ground motion simulation methods implemented on the SCEC Broadband Platform (BBP), version 14.3 (as of March 2014). A seven-member panel, the authorship of this article, was formed to evaluate those methods for the prediction of pseudo-‐spectral accelerations (PSAs) of ground motion. The panel’s mandate was to evaluate the methods using tools developed through the validation exercise (Goulet et al. ,2014), and to define validation metrics for the assessment of the methods’ performance. This paper summarizes the evaluation process and conclusions from the panel. The five broadband, finite-source simulation methods on the BBP include two deterministic approaches herein referred to as CSM (Anderson, 2014) and UCSB (Crempien and Archuleta, 2014); a band-‐limited stochastic white noise method called EXSIM (Atkinson and Assatourians, 2014); and two hybrid approaches, referred to as G&P (Graves and Pitarka, 2014) and SDSU (Olsen and Takedatsu, 2014), which utilize a deterministic Green’s function approach for periods longer than 1 second and stochastic methods for periods shorter than 1 second. Two acceptance tests were defined to validate the broadband finite‐source ground methods (Goulet et al., 2014). Part A compared observed and simulated PSAs for periods from 0.01 to 10 seconds for 12 moderate to large earthquakes located in California, Japan, and the eastern US. Part B compared the median simulated PSAs to published NGA-‐West1 (Abrahamson and Silva, 2008; Boore and Atkinson, 2008; Campbell and Bozorgnia, 2008; and Chiou and Youngs, 2008) ground motion prediction equations (GMPEs) for specific magnitude and distance cases using a pass-‐fail criteria based on a defined acceptable range around the spectral shape of the GMPEs. For the initial Part A and Part B validation exercises during the summer of 2013, the software for the five methods was locked in at version 13.6 (see Maechling et al., 2014). In the spring of 2014, additional moderate events were considered for the Part A validation, and additional magnitude and distance cases were considered for the Part B validation, for the software locked in at version 14.3. Several of the simulation procedures, specifically UCSB and SDSU, changed significantly between versions 13.6 and 14.3. The CSM code was not submitted in time for the v14.3 evaluation and its detailed performance is not addressed in this paper. As described in Goulet et al. (2014) and Maechling et al. (2014), the BBP generates a variety of products, including three-‐component acceleration time series. A series of post-‐processing codes were developed to provide individual component PSAs and average median horizontal-‐component PSA (referred to as RotD50; Boore, 2010) for oscillator periods ranging from 0.01 to 10 seconds, as well as median PSA values computed using the NGA-‐West 1 GMPEs. The BBP was also configured to provide statistical analysis of simulation results relative to recordings (Part A) and GMPEs (Part B) as described further in sections below. As part of our evaluation, we reviewed documentation provided by each of the developers, which included the technical basis behind the methods and the developer’s self-‐assessments regarding the extrapolation capabilities (in terms of magnitude and distance ranges) of their methods. Two workshops were held in which methods and results were presented, and the panel was given the opportunity to question the developers and to have detailed technical discussions. A SCEC report (Dreger et al., 2013) describes the results of this review for BBP version 13.6. This paper summarizes that work and presents results for the more recent BBP 14.3 validation.

Seismological Research Letters↗

USGS2014

Seismic‐wave attenuation determined from tectonic tremor in multiple subduction zones

Tectonic tremor provides a new source of observations that can be used to constrain the seismic attenuation parameter for ground‐motion prediction and hazard mapping. Traditionally, recorded earthquakes of magnitude ∼3–8 are used to develop ground‐motion prediction equations; however, typical earthquake records may be sparse in areas of high hazard. In this study, we constrain the distance decay of seismic waves using measurements of the amplitude decay of tectonic tremor, which is plentiful in some regions. Tectonic tremor occurs in the frequency band of interest for ground‐motion prediction (i.e., ∼2–8 Hz) and is located on the subducting plate interface, at the lower boundary of where future large earthquakes are expected. We empirically fit the distance decay of peak ground velocity from tremor to determine the attenuation parameter in four subduction zones: Nankai, Japan; Cascadia, United States–Canada; Jalisco, Mexico; and southern Chile. With the large amount of data available from tremor, we show that in the upper plate, the lower crust is less attenuating than the upper crust. We apply the same analysis to intraslab events in Nankai and show the possibility that waves traveling from deeper intraslab events experience more attenuation than those from the shallower tremor due to ray paths that pass through the subducting and highly attenuating oceanic crust. This suggests that high pore‐fluid pressure is present in the tremor source region. These differences imply that the attenuation parameter determined from intraslab earthquakes may underestimate ground motion for future large earthquakes on the plate interface.

Cascadia↗

USGS2013

Ground-motion prediction from tremor

The widespread occurrence of tremor, coupled with its frequency content and location, provides an exceptional opportunity to test and improve strong ground-motion attenuation relations for subduction zones. We characterize the amplitude of thousands of individual 5 min tremor events in Cascadia during three episodic tremor and slip events to constrain the distance decay of peak ground acceleration (PGA) and peak ground velocity (PGV). We determine the anelastic attenuation parameter for ground-motion prediction equations (GMPEs) to a distance of 150 km, which is sufficient to place important constraints on ground-motion decay. Tremor PGA and PGV show a distance decay that is similar to subduction-zone-specific GMPEs developed from both data and simulations; however, the massive amount of data present in the tremor observations should allow us to refine distance-amplitude attenuation relationships for use in hazard maps, and to search for regional variations and intrasubduction zone differences in ground-motion attenuation.

British Columbia;Oregon;Washington↗

USGS2013

Stable stress‐drop measurements and their variability: Implications for ground‐motion prediction

We estimate the arms‐stress drop, Graphic, (Hanks, 1979) using acceleration time records of 59 earthquakes from two earthquake sequences in eastern Honshu, Japan. These acceleration‐based static stress drops compare well to stress drops calculated for the same events by Baltay et al. (2011) using an empirical Green’s function (eGf) approach. This agreement supports the assumption that earthquake acceleration time histories in the bandwidth between the corner frequency and a maximum observed frequency can be considered white, Gaussian, noise. Although the Graphic is computationally simpler than the eGf‐based Graphic‐stress drop, and is used as the “stress parameter” to describe the earthquake source in ground‐motion prediction equations, we find that it only compares well to the Graphic at source‐station distances of ∼20 km or less because there is no consideration of whole‐path anelastic attenuation or scattering. In these circumstances, the correlation between the Graphic and Graphic is strong. Events with high and low stress drops obtained through the eGf method have similarly high and low Graphic. We find that the inter‐event standard deviation of stress drop, for the population of earthquakes considered, is similar for both methods, 0.40 for the Graphic method and 0.42 for the Graphic, in log10 units, provided we apply the ∼20 km distance restriction to Graphic. This indicates that the observed variability is inherent to the source, rather than attributable to uncertainties in stress‐drop estimates

BSSA↗

USGS2012

Have recent earthquakes exposed flaws in or misunderstandings of probabilistic seismic hazard analysis?

In a recent Opinion piece in these pages, Stein et al. (2011) offer a remarkable indictment of the methods, models, and results of probabilistic seismic hazard analysis (PSHA). The principal object of their concern is the PSHA map for Japan released by the Japan Headquarters for Earthquake Research Promotion (HERP), which is reproduced by Stein et al. (2011) as their Figure 1 and also here as our Figure 1. It shows the probability of exceedance (also referred to as the “hazard”) of the Japan Meteorological Agency (JMA) intensity 6–lower (JMA 6–) in Japan for the 30-year period beginning in January 2010. JMA 6– is an earthquake-damage intensity measure that is associated with fairly strong ground motion that can be damaging to well-built structures and is potentially destructive to poor construction (HERP, 2005, appendix 5). Reiterating Geller (2011, p. 408), Stein et al. (2011, p. 623) have this to say about Figure 1: The regions assessed as most dangerous are the zones of three hypothetical “scenario earthquakes” (Tokai, Tonankai, and Nankai; see map). However, since 1979, earthquakes that caused 10 or more fatalities in Japan actually occurred in places assigned a relatively low probability. This discrepancy—the latest in a string of negative results for the characteristic model and its cousin the seismic-gap model—strongly suggest that the hazard map and the methods used to produce it are flawed and should be discarded. Given the central role that PSHA now plays in seismic risk analysis, performance-based engineering, and design-basis ground motions, discarding PSHA would have important consequences. We are not persuaded by the arguments of Geller (2011) and Stein et al. (2011) for doing so because important misunderstandings about PSHA seem to have conditioned them. In the quotation above, for example, they have confused important differences between earthquake-occurrence observations and ground-motion hazard calculations.

Seismological Research Letters↗

USGS2012

Aftershocks halted by static stress shadows

Earthquakes impart static and dynamic stress changes to the surrounding crust. Sudden fault slip causes small but permanent—static—stress changes, and passing seismic waves cause large, but brief and oscillatory—dynamic—stress changes. Because both static and dynamic stresses can trigger earthquakes within several rupture dimensions of a mainshock, it has proven difficult to disentangle their contributions to the triggering process 1–3 . However, only dynamic stress can trigger earthquakes far from the source 4,5 , and only static stress can create stress shadows, where the stress and thus the seismicity rate in the shadow area drops following an earthquake 6–9 . Here we calculate the stress imparted by the magnitude 6.1 Joshua Tree and nearby magnitude 7.3 Landers earthquakes that occurred in California in April and June 1992, respectively, and measure seismicity through time. We show that, where the aftershock zone of the first earthquake was subjected to a static stress increase from the second, the seismicity rate jumped. In contrast, where the aftershock zone of the first earthquake fell under the stress shadow of the second and static stress dropped, seismicity shut down. The arrest of seismicity implies that static stress is a requisite element of spatial clustering of large earthquakes and should be a constituent of hazard assessment.

California↗

USGS2009

Deep low-frequency earthquakes in tremor localize to the plate interface in multiple subduction zones

Deep tremor under Shikoku, Japan, consists primarily, and perhaps entirely, of swarms of low-frequency earthquakes (LFEs) that occur as shear slip on the plate interface . Although tremor is observed at other plate boundaries, the lack of cataloged low-frequency earthquakes has precluded a similar conclusion about tremor in those locales. We use a network autocorrelation approach to detect and locate LFEs within tremor recorded at three subduction zones characterized by different thermal structures and levels of interplate seismicity: southwest Japan, northern Cascadia, and Costa Rica. In each case we find that LFEs are the primary constituent of tremor and that they locate on the deep continuation of the plate boundary. This suggests that tremor in these regions shares a common mechanism and that temperature is not the primary control on such activity.

Geophysical Research Letters↗

USGS2008

The 1906 San Francisco earthquake a century later: Introduction to the special section

The great 1906 San Francisco earthquake is perhaps the landmark event in the history of earthquake science. It began with a foreshock at 5:12 a.m. local time in the morning of 18 April 1906. Some 30 sec later, the main event initiated on the San Andreas fault, just off the San Francisco coast ( Lawson, 1908 ). Within 90 sec, nearly 480 km of the San Andreas fault ruptured (see Fig. 1 ), extending south to the northern end of the creeping section near San Juan Bautista and north to the terminus of the fault at the triple junction near Cape Mendocino ( Song et al. , 2008 ). As it ruptured, it generated powerful seismic waves over the entire rupture length and set in motion a chain of events that led to the destruction of most of San Francisco, the largest city of the western United States at the time. The earthquake occurred in the early days of instrumental seismology, which renders the data difficult to analyze, but our best estimate is that the moment magnitude was 7.9 ( Song et al. , 2008 ), about 26 times the size of the 1989 magnitude 6.9 Loma Prieta earthquake as measured by seismic moment ( Hanks and Krawinkler, 1991 ; Song et al. , 2008 ).

California↗

USGS2004

Optimizing correlation techniques for improved earthquake location

Earthquake location using relative arrival time measurements can lead to dramatically reduced location errors and a view of fault-zone processes with unprecedented detail. There are two principal reasons why this approach reduces location errors. The first is that the use of differenced arrival times to solve for the vector separation of earthquakes removes from the earthquake location problem much of the error due to unmodeled velocity structure. The second reason, on which we focus in this article, is that waveform cross correlation can substantially reduce measurement error. While cross correlation has long been used to determine relative arrival times with subsample precision, we extend correlation measurements to less similar waveforms, and we introduce a general quantitative means to assess when correlation data provide an improvement over catalog phase picks. We apply the technique to local earthquake data from the Calaveras Fault in northern California. Tests for an example streak of 243 earthquakes demonstrate that relative arrival times with normalized cross correlation coefficients as low as ∼70%, interevent separation distances as large as to 2 km, and magnitudes up to 3.5 as recorded on the Northern California Seismic Network are more precise than relative arrival times determined from catalog phase data. Also discussed are improvements made to the correlation technique itself. We find that for large time offsets, our implementation of time-domain cross correlation is often more robust and that it recovers more observations than the cross spectral approach. Longer time windows give better results than shorter ones. Finally, we explain how thresholds and empirical weighting functions may be derived to optimize the location procedure for any given region of interest, taking advantage of the respective strengths of diverse correlation and catalog phase data on different length scales.

Bulletin of the Seismological Society of America↗

USGS1996

The Loma Prieta, California, Earthquake of October 17, 1989: Main shock characteristics

The October 17, 1989, Loma Prieta, Calif., earthquake (0004:15.2 G.m.t. October 18; lat 37.036º N., long 121.883º W.; 19-km depth) had a local magnitude (M L ) of about 6.7, a surface-wave magnitude (M S ) of 7.1, a seismic moment of 2.2x10 19 N-m to 3.5x10 19 N-m, a source duration of 6 to 15 s, and an average stress drop of at least 50 bars. Slip occurred on a dipping fault surface about 35 km long and was largely confined to a depth of about 7 to 20 km. The slip vector had a large vertical component, and slip was distributed in two main regions situated northwest and southeast of the hypocenter. This slip distribution caused about half of the earthquake's energy to be focused toward the urbanized San Francisco Bay region, while the other half was focused toward the southeast. Had the rupture initiated at the southeast end of the aftershock zone, shaking in the bay region would have been both longer and stronger. These source parameters suggest that the earthquake was not a typical shallow San Andreas-type event but a deeper event on a different fault with a recurrence interval of many hundreds of years. Therefore, the potential for a damaging shallow event on the San Andreas fault in the Santa Cruz Mountains may still exist.

California↗