Advertise on podcast: Earthquake Science Center Seminars

Emily Mongold, Stanford University The impact of liquefaction on a regional scale is not well understood or modeled with traditional approaches. This paper presents a method to quantitatively assess liquefaction hazard and risk on a regional scale, accounting for uncertainties in soil properties, groundwater conditions, ground shaking parameters, and empirical liquefaction potential index (LPI) equations. The regional analysis is applied to a case study to calculate regional occurrence rates for the extent and severity of liquefaction and to quantify losses resulting from ground shaking and liquefaction damage to residential buildings. We present a regional-scale metric to quantify the extent and severity of liquefaction. A sensitivity analysis on epistemic uncertainty indicates that the two most important factors on output liquefaction maps are the empirical liquefaction equation, emphasizing the necessity of incorporating multiple equations in future regional studies, and the water table level, highlighting concerns around data availability and sea level rise. Furthermore, the disaggregation of seismic sources reveals that triggering earthquakes for various extents of liquefaction originate from multiple sources, though primarily nearby faults and large magnitude ruptures. This finding indicates the value of adopting regional probabilistic analysis in future studies to capture the diverse sources and spatial distribution of liquefaction.

(1) Susan Hough, (2) Kate Hutton, (1) U.S. Geological Survey, (2) Caltech (retired) On the 30th anniversary of the 17 January 1994 Northridge, California, earthquake, we present a retrospective overview of an earthquake that had an enormous, multi-faceted impact in the greater Los Angeles area. In this two-part seminar, retired Caltech seismologist Kate Hutton first discusses the response to the earthquake by the (then) Southern California Seismic Network, which found itself slammed by joint demands of data analysis and overwhelming media/public interest. In the second part, Susan Hough discusses how modern analysis and data products – the development of which was spurred by the earthquake – bring early characterizations of the sequence, including its near-field ground motions, into greater focus.

Sarah Minson, U.S. Geological Survey There are many underdetermined geophysical inverse problems. For example, when we try to infer earthquake fault slip, we find that there are many potential slip models that are consistent with our observations and our understanding of earthquake physics. One way to approach these problems is to use Bayesian analysis to infer the ensemble of all potential models that satisfy the observations and our prior knowledge. In Bayesian analysis, our prior knowledge is known as the prior probability density function or prior PDF, the fit to the data is the data likelihood function, and the target PDF that satisfies both the prior PDF and data likelihood function is the posterior PDF. Simulating a posterior PDF can be computationally expensive. Typical earthquake rupture models with 10 km spatial resolution can require using Markov Chain Monte Carlo (MCMC) to draw tens of billions of random realizations of fault slip. And now new technological advancements like LiDAR provide enormous numbers of laser point returns that image surface deformation at submeter scale, exponentially increasing computational cost. How can we make MCMC sampling efficient enough to simulate fault slip distributions at sub-meter scale using “Big Data”? We present a new MCMC approach called cross-fading in which we transition from an analytical posterior PDF (obtained from a conjugate prior to the data likelihood function) to the desired target posterior PDF by bringing in our physical constraints and removing the conjugate prior. This approach has two key efficiencies. First, the starting PDF is by construction “close” to the target posterior PDF, requiring very little MCMC to update the samples to match the target. Second, all PDFs are defined in model space, not data space. The forward model and data misfit are never evaluated during sampling, allowing models to be fit to Big Data with zero computational cost. It is even possible, without additional computational cost, to incorporate model prediction errors for Big Data, that is, to quantify the effects on data prediction of uncertainties in the model design. While we present earthquake models, this approach is flexible and can be applied to many geophysical problems.

(1) Stephen Wu, (2) Keisuke Yano, Institute of Statistical Mathematics, Japan (1) Since 2021, the Seismology TowArd Research innovation with data of Earthquake (STAR-E) project has been established by the Japanese government to promote interdisciplinary research between data science and seismology. Five proposals have been accepted to be the core projects of STAR-E and EEW has become a sub-project in one of the selected projects. In this talk, I will provide an overview of the plan to improve EEW in Japan through integration with data science. While no concrete results have been obtained yet, I will share part of the blueprint of the possible EEW development in Japan for the future 5 years. (2) In real data analysis, we often encounter mixed-domain data. Mixed-domain data refer to multivariate data in various domains such as real values, categorical values, manifold values, and functional values. In this presentation, we will introduce our minimum information dependence model. This statistical model is tailored to analyze mixed-domain data with potential higher-order dependencies. We will highlight its utility through its application in the ecological study of penguins and in the analysis of earthquake catalogs.

(1) Rachel Abercrombie, (2) Annemarie Baltay, (1) Boston University, (2) U.S. Geological Survey In 2021 we launched the Community Stress Drop Validation Study, focused on the 2019 Ridgecrest earthquake, California, sequence, using a common dataset. The broad aim of the collaboration is to improve the quality of estimates of stress drop and related fundamental earthquake source parameters (corner frequency, source duration, etc.) and their uncertainties, to enable more reliable ground motion forecasting, and to obtain a better understanding of earthquake source physics. Seismological estimates of stress drop from earthquake spectral measurements have become standard practice over the last 50 years, but their wide variability, model dependence and inconsistency between studies have led to controversy and concerns about how to assess and interpret these measurements. The SCEC/USGS community study has engaged a wide international community focused on improving methods and distinguishing the sources of variability between physical earthquake source variation, and random and systematic scatter and bias. To date, 18 research groups have submitted 28 different measurements of source parameters for earthquakes in the 2019 Ridgecrest sequence, with a focus on 55 events of M2 to 5. These approaches include spectral decomposition/generalized inversion, empirical Green’s function analysis in both frequency and time domains, and ground-motion and single-station based approaches. Comparison of submitted stress drops reveals considerable systematic and random scatter, but also shows consistency between events; for some events, methods are in agreement on either relatively high or low stress drops. Ongoing focus is on understanding the relative influences of different analysis parameter choices, assumptions about attenuation, frequency range of the data, and the growing evidence of widespread complexity and heterogeneity in even small earthquake ruptures. We welcome new members wishing to observe, learn or more actively participate; more information can be found at https://www.scec.org/research/stress-drop-validation.

Heather Crume, California Geological Survey Surface creep has been documented on the San Andreas fault (SAF) since the 1960s. From Parkfield in the southeast to San Juan Bautista (SJB) in the northwest, the SAF is largely creeping and accommodating most of the ~38 mm/year right-lateral plate motion. The SJB section of the SAF lies at the northwest boundary of the central creeping section, forming a creeping-to-locked transition. These transition sections are known to be potential zones for earthquake nucleation. Spatiotemporal changes in fault creep within this locking transition provide a potential quantitative measure for the assessment of the seismic hazard of the SAF system in the region. In addition to steady fault creep, episodic creep events characterized by accelerated slip of a few millimeters to centimeters over several days occur. However, knowledge of the along-strike and downdip-extent and propagation velocity of these events is limited by the sparse density of current creepmeters. How do these events propagate? What is their magnitude? What factors drive their occurrence? Moreover, there is a need to address environmental effects that can confound creep data and develop appropriate corrections. To address some of these questions, we have initiated a densification of the current creepmeter array by installing new instruments and renovating those in disrepair. We report results from some of these sensors. At Fox Creek, south of Hollister, we installed two creepmeters 130 m apart to measure creep event propagation velocity. One was equipped with an orthogonal sensor to measure dilation. During the dry season ≈0.16-0.35 mm of fault dilation accompanied complex creep event sequences with cumulative amplitudes of 3.5-5.8 mm. In the months following each sequence the fault zone slowly returned to its pre-event width. During one creep event a southward propagating dislocation was present with a velocity of 0.4 km/hour. We also observe distinct differences in amplitude and shape of creep events. Further, we are able to correct for apparent left-lateral slip due to fault closing during a rain event using the relationship between the orthogonal and oblique instruments.

Shuo Ma, San Diego State University One crucial yet unanswered question about the 2011 Tohoku-Oki earthquake and tsunami is what generated the largest tsunami (up to 40 m) along the Sanriku coast north of 39°N without large slip near the trench. A minimalist dynamic rupture model with wedge plasticity is presented to address this issue. The model incorporates the important variation of sediment thickness along the Japan Trench into the Japan Integrated Velocity Structure Model (JIVSM). By revising a heterogeneous stress drop model, the dynamic rupture model with a standard rate-and-state friction law can well explain the GPS, tsunami, and differential bathymetry data (within data uncertainties) with minimum model tuning. The rupture is driven by a large patch of stress drop up to ~10 MPa near the hypocenter with significantly smaller stress drop ( 3 MPa) in the upper ~10 km. The largest shallow slip reaches 75.67 m close to the trench ~50 km north of hypocenter dominated by elastic off-fault response, which is caused by the large fault width, free surface, shallowly dipping fault geometry, and increasing sediment thickness northward. North of large shallow slip zone, however, inelastic deformation of thick wedge sediments significantly controls the rupture propagation along trench, giving rise to slow rupture velocity (~850 m/s), diminishing shallow slip, and efficient seafloor uplift. The short-wavelength inelastic uplift produces impulsive tsunami consistent with the observations off the Sanriku coast in terms of timing, amplitude, and pulse width. Wedge plasticity and variation of sediment thickness along the Japan Trench thus provides a self-consistent interpretation to the along-strike variation of near-trench slip and anomalous tsunami generation in the northern Japan Trench in this earthquake.

Artemii Novoselov, Stanford University This seminar introduces PhaseHunter, a deep learning framework initially designed for the precise estimation and uncertainty quantification of seismic phase onset times. Building upon this foundational capability, PhaseHunter has evolved to handle a broader range of seismic applications through a probabilistic deep learning regression approach. This enables the framework to analyze both continuous and binary properties of seismic signals, thereby extending its potential applications to include earthquake location, seismic tomography, source discrimination, and earthquake early warning systems. The seminar will explore the technical aspects and practical applications of PhaseHunter, offering insights into how this tool could serve various facets of seismological research and hazard assessment. As an open-source project, PhaseHunter also encourages community contributions for ongoing improvements and adaptations.

James Neely, University of Chicago Commonly used large earthquake recurrence models have two major limitations. First, they predict that the probability of a large earthquake stays constant or even decreases after it is “overdue” (past the observed average recurrence interval), so additional accumulated strain does not make an earthquake more likely. Second, they assume that the probability distribution of the time between earthquakes is the same over successive earthquake cycles, despite the fact that earthquake histories show clusters and gaps. These limitations arise because the models are purely statistical in that they do not incorporate fundamental aspects of the strain accumulation and release processes that cause earthquakes. Here, we present a new large earthquake probability model, built on the Long-Term Fault Memory model framework that better reflects the strain accumulation and release processes. This Generalized Long-Term Fault Memory model (GLTFM) assumes that earthquake probability always increases with time between earthquakes as strain accumulates and allows the possibility of earthquakes releasing only part of the strain accumulated on the fault. GLTFM estimates when residual strain is likely present and its impact on the probable timing of the next earthquake in the sequence, and so can describe clustered earthquake sequences better than commonly used models. GLTFM’s simple implementation and versatility should make it a powerful tool in earthquake forecasting.

Zhiang Chen, California Institute of Technology The intricate and dynamic nature of fault zones and fragile geological features has long fascinated geoscientists and researchers. Understanding these geological phenomena is crucial not only for scientific exploration but also for hazard assessment and resource management. Recently, the convergence of robotics and machine learning has given rise to a transformative practice called automated geoscience. This practice utilizes robotics to automate data collection and machine learning to automate data processing, liberating geoscientists from labor-intensive activities. Focusing on rock detection, mapping, and dynamics analysis, I present the applications of automated geoscience in fault zone mapping and fragile geological feature analysis. To explore the influence of rocky fault scarp development on rock trait distributions, I have developed a data-processing pipeline using UAVs and deep learning to segment dense rock distributions. This application provides a statistical approach for geomorphology studies. Precariously balanced rocks (PBRs) offer insights into ground motion constraints for hazard analysis. I have designed offboard and onboard methods for autonomous PBR detection and mapping. After mapping, I delve into PBR dynamics with a virtual shake robot simulating the dynamics of PBRs in overturning and large displacement processes with respect to various ground motions. The overturning and large displacement processes provide upper-bound and lower-bound ground motion constraints, respectively. Moving forward, I am integrating automated geoscience into broader studies on fault zone mapping and fragile geological feature analysis. My aim is to push this interdisciplinary research direction, offering potential advancements in hazard monitoring and prospecting.