Showing papers in "Bulletin of the Seismological Society of America in 2018"

Generalized Seismic Phase Detection with Deep Learning

Abstract: To optimally monitor earthquake‐generating processes, seismologists have sought to lower detection sensitivities ever since instrumental seismic networks were started about a century ago. Recently, it has become possible to search continuous waveform archives for replicas of previously recorded events (i.e., template matching), which has led to at least an order of magnitude increase in the number of detected earthquakes and greatly sharpened our view of geological structures. Earthquake catalogs produced in this fashion, however, are heavily biased in that they are completely blind to events for which no templates are available, such as in previously quiet regions or for very large‐magnitude events. Here, we show that with deep learning, we can overcome such biases without sacrificing detection sensitivity. We trained a convolutional neural network (ConvNet) on the vast hand‐labeled data archives of the Southern California Seismic Network to detect seismic body‐wave phases. We show that the ConvNet is extremely sensitive and robust in detecting phases even when masked by high background noise and when the ConvNet is applied to new data that are not represented in the training set (in particular, very large‐magnitude events). This generalized phase detection framework will significantly improve earthquake monitoring and catalogs, which form the underlying basis for a wide range of basic and applied seismological research.

Surface Rupture of Multiple Crustal Faults in the 2016 Mw 7.8 Kaikōura, New Zealand, Earthquake

Abstract: Multiple (>20 >20 ) crustal faults ruptured to the ground surface and seafloor in the 14 November 2016 M w Mw 7.8 Kaikōura earthquake, and many have been documented in detail, providing an opportunity to understand the factors controlling multifault ruptures, including the role of the subduction interface. We present a summary of the surface ruptures, as well as previous knowledge including paleoseismic data, and use these data and a 3D geological model to calculate cumulative geological moment magnitudes (M G w MwG ) and seismic moments for comparison with those from geophysical datasets. The earthquake ruptured faults with a wide range of orientations, sense of movement, slip rates, and recurrence intervals, and crossed a tectonic domain boundary, the Hope fault. The maximum net surface displacement was ∼12 m ∼12 m on the Kekerengu and the Papatea faults, and average displacements for the major faults were 0.7–1.5 m south of the Hope fault, and 5.5–6.4 m to the north. M G w MwG using two different methods are M G w MwG 7.7 +0.3 −0.2 7.7−0.2+0.3 and the seismic moment is 33%–67% of geophysical datasets. However, these are minimum values and a best estimate M G w MwG incorporating probable larger slip at depth, a 20 km seismogenic depth, and likely listric geometry is M G w MwG 7.8±0.2 7.8±0.2 , suggests ≤32% ≤32% of the moment may be attributed to slip on the subduction interface and/or a midcrustal detachment. Likely factors contributing to multifault rupture in the Kaikōura earthquake include (1) the presence of the subduction interface, (2) physical linkages between faults, (3) rupture of geologically immature faults in the south, and (4) inherited geological structure. The estimated recurrence interval for the Kaikōura earthquake is ≥5,000–10,000 yrs ≥5,000–10,000 yrs , and so it is a relatively rare event. Nevertheless, these findings support the need for continued advances in seismic hazard modeling to ensure that they incorporate multifault ruptures that cross tectonic domain boundaries.

Spatial and Spectral Interpolation of Ground-Motion Intensity Measure Observations

Abstract: Following a significant earthquake, ground-motion observations are available for a limited set of locations and intensity measures (IMs). Typically, however, it is desirable to know the ground motions for additional IMs and at locations where observations are unavailable. Various interpolation methods are available, but because IMs or their logarithms are normally distributed, spatially correlated, and correlated with each other at a given location, it is possible to apply the conditional multivariate normal (MVN) distribution to the problem of estimating unobserved IMs. In this article, we review the MVN and its application to general estimation problems, and then apply the MVN to the specific problem of ground-motion IM interpolation. In particular, we present (1) a formulation of the MVN for the simultaneous interpolation of IMs across space and IM type (most commonly, spectral response at different oscillator periods) and (2) the inclusion of uncertain observation data in the MVN formulation. These techniques, in combination with modern empirical ground-motion models and correlation functions, provide a flexible framework for estimating a variety of IMs at arbitrary locations. Electronic Supplement: Demonstration Python script for the evaluation of the multivariate normal (MVN) with additional uncertainty.

Strong Correlation between Stress Drop and Peak Ground Acceleration for Recent M 1-4 Earthquakes in the San Francisco Bay Area

Abstract: Theoretical and observational studies suggest that between-event variability in the median ground motions of larger (M ≥ 5) earthquakes is controlled primarily by the dynamic properties of the earthquake source, such as Brune-type stress drop. Analogous results remain equivocal for smaller events due to the lack of comprehensive and overlapping ground-motion and source-parameter datasets in this regime. Here, we investigate the relationship between peak ground acceleration (PGA) and dynamic stress drop for a new dataset of 5297 earthquakes that occurred in the San Francisco Bay area from 2002 through 2016. For each event, we measure PGA on horizontal-component channels of stations within 100 km and estimate stress drop from P-wave spectra recorded on vertical-component channels of the same stations. We then develop a nonparametric ground-motion prediction equation (GMPE) applicable for the moderate (M 1–4) earthquakes in our study region, using a mixed-effects generalization of the Random Forest algorithm. We use the Random Forest GMPE to model the joint influence of magnitude, distance, and near-site effects on observed PGA. We observe a strong correlation between dynamic stress drop and the residual PGA of each event, with the events with higher-than-expected PGA associated with higher values of stress drop. The strength of this correlation increases as a function of magnitude but remains significant even for smaller magnitude events with corner frequencies that approach the observable bandwidth of the acceleration records. Mainshock events are characterized by systematically higher stress drop and PGA than aftershocks of equivalent magnitude. Coherent local variations in the distribution of dynamic stress drop provide observational constraints to support the future development of nonergodic GMPEs that account for variations in median stress drop at different source locations. Electronic Supplement: Figures showing the relation between Mw and ML, comparison of the ground-motion measurements from this study with the cross-listed records in the Next Generation Attenuation ground-motion database, the validation curve used to select the optimal tree depth for the Random Forest ground-motion prediction equation (GMPE) used in this study, the between-event ground-motion residual is plotted versus: (a) stress drop, (b) magnitude-adjusted stress drop, (c) depth, and (d) depth-adjusted stress drop, a table containing the ground-motion and stressdrop measurements associated with this study, and an example Python notebook.

Prenolin: International benchmark on 1D nonlinear: Site-response analysis—validation phase exercise

Abstract: This article presents the main results of the validation phase of the PRENOLIN project. PRENOLIN is an international benchmark on 1D nonlinear (NL) site‐response analysis. This project involved 19 teams with 23 different codes tested. It was divided into two phases; with the first phase verifying the numerical solution of these codes on idealized soil profiles using simple signals and real seismic records. The second phase described in this article referred to code validation for the analysis of real instrumented sites. This validation phase was performed on two sites (KSRH10 and Sendai) of the Japanese strong‐motion networks KiK‐net and Port and Airport Research Institute (PARI), respectively, with a pair of accelerometers at surface and depth. Extensive additional site characterizations were performed at both sites involving in situ and laboratory measurements of the soil properties. At each site, sets of input motions were selected to represent different peak ground acceleration (PGA) and frequency content. It was found that the code‐to‐code variability given by the standard deviation of the computed surface‐response spectra is around 0.1 (in log10 scale) regardless of the site and input motions. This indicates a quite large influence of the numerical methods on site‐effect assessment and more generally on seismic hazard. Besides, it was observed that site‐specific measurements are of primary importance for defining the input data in site‐response analysis. The NL parameters obtained from the laboratory measurements should be compared with curves coming from the literature. Finally, the lessons learned from this exercise are synthesized, resulting also in a few recommendations for future benchmarking studies, and the use of 1D NL, total stress site‐response analysis.

The 2013-2016 Induced Earthquakes in Harper and Sumner Counties, Southern Kansas

Abstract: We examine the first four years (2013–2016) of the ongoing seismicity in southern Kansas using high-precision locations derived from a local seismometer network. The earthquakes occur almost exclusively in the shallow crystalline basement, below the wastewater injection horizon of the Arbuckle Group at the base of the sedimentary section. Multiple lines of evidence lead us to conclude that disposal of wastewater from the production of oil and gas by deep injection is the probable cause for the surge of seismicity that began in 2013. First, the seismicity correlates in space and time with the injection. We observe increases in seismicity subsequent to increases in injection and decreases in seismicity in response to decreases in injection. Second, the earthquake-rate change is statistically improbable to be of natural origin. From 1974 through the time of the injection increase in 2012, no ML 4 or larger earthquakes occurred in the study area, while six occurred between 2012 and 2016. The probability of this rate change occurring randomly is ∼0:16%. Third, the other potential industrial drivers of seismicity (hydraulic fracturing and oil production) do not correlate in space or time with seismicity. Local geological conditions are important in determining whether injection operations will induce seismicity, as shown by absence of seismicity near the largest injection operations in the southwest portion of our study area. In addition to local operations, the presence of seismicity 10+ km from large injection wells indicates that regional injection operations also need to be considered to understand the effects of injection on seismicity. Electronic Supplement: Tables of seismic stations used, and earthquake catalogs and figures showing oil and gas operational histories and their relationship to seismicity.

Sources of Long-Range Anthropogenic Noise in Southern California and Implications for Tectonic Tremor Detection

Abstract: We study anthropogenic noise sources seen on seismic recordings along the central section of the San Jacinto fault near Anza, southern California. The strongest signals are caused by freight trains passing through the Coachella Valley north of Anza. Train-induced transients are observed at distances of up to 50 km from the railway, with durations of up to 20 min, and spectra that are peaked between 3 and 5 Hz. Additionally, truck traffic through the Coachella Valley generates a sustained hum with a similar spectral signature as the train transients but with lower amplitude. We also find that wind turbine activity in northern Baja California introduces a seasonal modulation of 1– to 5-Hz energy across the Anza network. We show that the observed train-generated transients can be used to constrain shallow attenuation structure at Anza. Using the results from this study as well as available borehole data, we further evaluate the performance of approaches that have been used to detect nonvolcanic tremor at Anza. We conclude that signals previously identified as spontaneous tremor (Hutchison and Ghosh, 2017) were probably generated by other nontectonic sources such as trains. Electronic Supplement: Seismograms during a transient, decomposition of the large amplitude signal into sequences of repeating patterns, detection statistics in the two stages of the tremor search procedure, velocity amplitude spectra of transients from two collocated borehole–surface station pairs, median weekday–night spectra for all stations in the Plate Boundary Observatory (PBO) Anza network, time variations of root mean square (rms) velocity for all stations in the PBO Anza network, cumulative population and kilometers of railroad on which the allowed speed for freight trains exceeds 50 mph, envelopes of vertical ground motions as a function of time, vertical ground motions recorded by the Anza borehole stations, and spectrograms of vertical ground motion and corresponding wind speeds.

A New Seismic Hazard Model for Ecuador

Abstract: We present a comprehensive probabilistic seismic hazard study for Ecuador, a country exposed to a high seismic hazard from megathrust subduction earthquakes and moderate-to-large shallow crustal earthquakes. Building on knowledge gained during the last decade about historical and contemporary seismicity, active tectonics, geodynamics, and geodesy, several alternative earthquake recurrence models have been developed. We propose an areal seismic zonation for the seismogenic crustal, inslab, and interface sources, modified from Yepes et al. (2016), to account for the information gained after the 2016Mw 7.8 Pedernales megathrust earthquake. Three different earthquake catalogs are used to account for uncertainties in magnitude–frequency distribution modeling. This first approach results in low hazard estimates for some areas near active crustal fault systems with low instrumental seismicity, but where geology and/or geodesy document rapid slip rates and high seismic potential. Consequently, we develop an alternative fault and background model that includes faults with earthquake recurrence models inferred from geologic and/or geodetic slip-rate estimates. The geodetic slip rates for a set of simplified faults are estimated from a Global Positioning System (GPS) horizontal velocity field from Nocquet et al. (2014). Various scenarios are derived by varying the percentage of motion that takes place aseismically. Combining these alternative earthquake recurrence models in a logic tree, and using a set of selected ground-motion models adapted to Ecuador’s different tectonic settings, mean hazard maps are obtained with their associated uncertainties. At the sites where uncertainties on hazard estimates are highest (difference between 84th and 16th percentiles > 0:4g), the overall uncertainty is controlled by the epistemic uncertainty on the source model.

Ground motions from the 7 and 19 September, 2017 Tehuantepec and Puebla-Morelos, Mexico, earthquakes

Abstract: The 2017 M 8.2 Tehuantepec and M 7.1 Puebla-Morelos earthquakes were deep inslab normal-faulting events that caused significant damage to several central-to-southern regions of Mexico. Inslab earthquakes are an important component of seismicity and seismic hazard in Mexico. Ground-motion prediction equations (GMPEs) are an integral part of seismic hazard assessment as well as risk and rapid-response products. This work examines the observed ground motions from these two events in comparison to the predicted median ground motions from four GMPEs. The residuals between the observed and modeled ground motions allow us to study regional differences in shaking, the effects of each earthquake, and basin effects in Mexico City, Puebla, and Oaxaca. We find that the ground motions from these two earthquakes are generally well modeled by the GMPEs. However, the Tehuantepec event shows larger than expected ground motions at greater distances and longer periods, which suggests a waveguide effect from the subduction zone geometry. Finally, Mexico City and the cities of Puebla and Oaxaca exhibit very large ground motions, indicative of well-known site and basin effects that are much stronger than the basin terms included in some of the GMPEs. Simple and rapid ground-motion parameter estimates that include site effects are key for hazard and real-time risk assessments in regions such as Mexico, where the vast majority of the population lives in areas where the aforementioned effects are relevant. However, GMPEs based on site correction terms dependent on topographic slope proxies underestimate, at least in the three cities tackled in this work, the observed amplification. Therefore, there is a need to improve models of seismic amplification in basins that could be included in GMPEs. Electronic Supplement: Tables of ground-motion intensity measures for each station and earthquake, as well as the residual uncertainties for each model, over all distances, and figures showing comprehensive ground-motion prediction equation (GMPE) and residuals results, for every period considered in this study, and the uncertainties.

Coseismic Rupture and Preliminary Slip Estimates for the Papatea Fault and Its Role in the 2016 Mw 7.8 Kaikōura, New Zealand, Earthquake

Abstract: Coseismic rupture of the 19‐km‐long north‐striking and west‐dipping sinistral reverse Papatea fault and nearby structures and uplift/translation of the Papatea block are two of the exceptional components of the 14 November 2016 Mw 7.8 Kaikōura earthquake. The dual‐stranded Papatea fault, comprising main (sinistral reverse) and western (dip‐slip) strands, ruptured onshore and offshore from south of Waipapa Bay to George Stream in the north, bounding the eastern side of the Papatea block. Fault rupture mapping was aided by the acquisition of multibeam bathymetry, light detection and ranging (lidar) topography and other imagery, as well as differential lidar (D‐lidar) from along the coast and Clarence River valley. On land, vertical throw and sinistral offset on the Papatea fault was assessed across an aperture of ±100 m using uncorrected D‐lidar and field data to develop preliminary slip distributions. The maximum up‐to‐the‐west throw on the main strand is ∼9.5±0.5 m⁠, and the mean throw across the Papatea fault is ∼4.5±0.3 m⁠. The maximum sinistral offset, measured near the coast on the main strand, is ∼6.1±0.5 m⁠. From these data, and considering fault dip, we calculate a maximum net slip of 11.5±2 m and an average net slip of 6.4±0.2 m for the Papatea fault surface rupture in 2016. Large sinistral reverse displacement on the Papatea fault is consistent with uplift and southward escape of the Papatea block as observed from Interferometric Synthetic Aperture Radar (InSAR) and optical image correlation datasets. The throw and net slip are exceedingly high for the length of the Papatea fault; such large movements likely only occur during multifault Kaikōura‐type earthquakes that conceivably have recurrence times of ≥5000–12,000 yrs⁠. The role of the Papatea fault in the Kaikōura earthquake has significant implications for characterizing complex fault sources in seismic hazard models.

Decomposing Leftovers: Event, Path, and Site Residuals for a Small‐Magnitude Anza Region GMPE

Abstract: Ground-motion prediction equations (GMPEs) are critical elements of probabilistic seismic hazard analysis (PSHA), as well as for other applications of ground motions. To isolate the path component for the purpose of building nonergodic GMPEs, we compute a regional GMPE using a large dataset of peak ground accelerations (PGAs) from small-magnitude earthquakes (0:5 ≤ M ≤ 4:5 with >10; 000 events, yielding ∼120; 000 recordings) that occurred in 2013 centered around the ANZA seismic network (hypocentral distances ≤180 km) in southern California. We examine two separate methods of obtaining residuals from the observed and predicted ground motions: a pooled ordinary least-squares model and a mixed-effects maximum-likelihood model. Whereas the former is often used by the broader seismological community, the latter is widely used by the ground-motion and engineering seismology community. We confirm that mixed-effects models are the preferred and most statistically robust method to obtain event, path, and site residuals and discuss the reasoning behind this. Our results show that these methods yield different consequences for the uncertainty of the residuals, particularly for the event residuals. Finally, our results show no correlation (correlation coefficient [CC] <0:03) between site residuals and the classic site-characterization term VS30, the time-averaged shearwave velocity in the top 30 m at a site. We propose that this is due to the relative homogeneity of the site response in the region and perhaps due to shortcomings in the formulation of VS30 and suggest applying the provided PGA site correction terms to future ground-motion studies for increased accuracy. Electronic Supplement: Peak ground acceleration (PGA) dataset.