Earthquake Science Center Seminars

During the 2018 eruption of Kilauea, the central part of the caldera collapsed by up to 500 meters and caused M~5 earthquakes, almost daily from the end of May until the beginning of August. These earthquakes damaged the USGS Hawaiian Volcano Observatory and future events present a hazard to other structures around the caldera. But how can we assess the hazard of these highly clustered earthquakes, which do not follow a Poisson process? After all, in his seminal paper on Probabilistic Seismic Hazard Assessment (PSHA), Allin Cornell wrote, "the assumption that the occurrences of earthquakes follow the behavior of the Poisson process model can be removed only at a great penalty" (BSSA, 1968). I will discuss a pair of papers by Andrea Llenos and myself that address this issue. In the first, we model the occurrence of these earthquakes by combining a distribution for the occurrence of the underlying process of caldera collapses with a distribution for the earthquake response to those collapses. This approach could also be used to model the hazard due to aftershocks, which can be considered the earthquake response to the underlying process of mainshocks. While simulations can be used to compute seismic hazard including behavior such as aftershocks (e.g. Field et al., BSSA, 2017), the computational requirements may be daunting for many applications. A clear goal of Cornell was an analytic method to allow users to explore the impact of a variety of parameters. Therefore, our second paper introduces an analytic approach for PSHA that allows for arbitrary distributions of earthquake occurrence without an appreciable computational penalty. We apply this method to the caldera collapse earthquakes and explore general implications of non-Poissonian behavior for PSHA.

If most of continental earthquakes occur in the first 15-20 km of the upper crust, lower crust earthquakes exist at depths of 50-60 km in the roots of mountain chains. Such earthquakes are observed, for example, below the Himalayas. Three mechanisms may explain how elastic strain energy is released in the form of earthquakes in the lower crust where rocks are often inferred to be ductile. The first mechanism involves local stress concentrations related to displacements along shear zones. The second mechanism involves the triggering of lower crust earthquakes by strain pulses produced by upper crust earthquakes. The third mechanism involves hydration reaction that may cause embrittlement of the rock. In Norway, outcrops of lower crust have recorded earthquake activity that occurred more than 400 million years ago during the Caledonian orogeny and that allow to test these three mechanisms. Fossil lower crust earthquake are observed because coseismic frictional heating produced numerous pseudotachylytes and fault offsets of up to two meters. Field data at the outcrop scale and images of fault samples in 2D (electron microscopy) and 3D (synchrotron X-ray microtomography) reveal the anatomy of the fault zone. Seismic slip and associated melting are preceded by fracturing, asymmetric fragmentation, and comminution of the wall rock caused by a dynamically propagating rupture. Data also show that earthquake activity preceded or was coeval with shear zone development and that fluids infiltrated into the earthquake damage zones. These observations indicate that, under mountain chains, the lower continental crust can be initially strong and not as ductile as usually proposed. The damage produced by earthquakes may explain how the rock rheology may evolve from brittle (e.g. granulites) to ductile (e.g. eclogites) through fluid infiltration.

This seminar will focus on the programs in the Office of Science, Quality, and Integrity (OSQI) that provide education, training, and opportunities for emerging and early career scientists. This includes internships and outreach opportunities in the Youth and Education in Science (YES) office, as well as programs that engage diverse and early career scientists. It will highlight resources and best practices for USGS personnel who wish to inclusively engage youth.

Eleanour Snow is the USGS Manager of Youth and Education programs, and has been working in the areas of Justice, Equity, Diversity, and Inclusion for decades. She is currently the co-lead of the Federal Interagency Working Group for Inclusion in STEM, and leader of the DOI Interagency Youth Team.

South-central Alaska is a region of high seismicity that is frequently impacted by intraslab earthquakes. Intraslab earthquakes, which occur at mid-crustal depths, do not produce the characteristic surficial faulting, land-level change, or tsunami deposits typically associated with megathrust earthquakes, but instead generate a more subtle signal in the geologic record. The emerging field of lacustrine and fjord paleoseismology utilizes relatively small and well-defined basins with unique depositional characteristics (e.g., varve formation), in combination with earthquake-generated turbidity deposits as paleoseismic proxies, to construct earthquake histories that are sensitive enough to record intraslab events. The 2018 Anchorage Earthquake resulted in high-intensity shaking across the upper Cook Inlet and highlighted the region's vulnerability to intraslab earthquakes. Recent work on Eklutna Lake by Van Daele et al. (2020) has confirmed the presence of earthquake-generated turbidites as a result of the 2018 earthquake across the lake's two sub-basins. The results of their study provided some of first inputs necessary to begin calibrating south-central Alaska's natural seismograph. However, important questions remain, including the minimum MMI necessary to trigger turbidity currents, which depositional environments are most susceptible to failure by seismic triggers, and the potential of relative turbidite thickness as an indicator of epicentral direction. I'll present preliminary results from fieldwork carried out this summer 2020 in south-central Alaska that focused on the collection of short-barrel gravity cores, MCS Sparker profiles, and high-resolution Chirp data. The goal of this project will be to characterize the sedimentary response of south-central Alaska's lakes and fjords to the 2018 Anchorage Earthquake. As a first step, short-barrel gravity cores are used to investigate the spatial extent of turbidity currents triggered by the 2018 Anchorage Earthquake, and if possible, identify a correlating acoustic signature.

The detailed examination of an earthquake's rupture is often reserved for the largest events, while smaller events are often characterized using methods that assume they have simple ruptures. This oversimplification allows us to obtain rupture characteristics for large numbers of earthquakes, but loses all information about complexity in the ruptures. Using multiple methods to characterize earthquake ruptures we explore the biases and limitations of the methods that assume a simple rupture and the benefits of examining heterogeneous ruptures of moderate (4-6M) to micro earthquakes (<3M). We examine the most common method of rupture characterization, the use of an event's spectrum and a simple model to estimate its rupture area and stress drop. Using a wide variety of data and methods to retrieve an event's source spectrum we find stress drop values are systematically biased depending on how site and path effects are removed, the wave type used (P-wave, S-wave), and the window length used to select the data. However, the relative stress drop values and the spatial and temporal interpretations of stress drops are consistent across methods. We then examine alternative methods of event characterization that do not assume an event's rupture is simple to determine if smaller events exhibit rupture complexity as observed in larger events. By utilizing finite fault slip distributions and source time function analysis we find that moderate and microearthquakes often exhibit complex multiphase ruptures. The factors that influence these more heterogeneous ruptures vary with region. Prior seismicity, fault maturity, and fault structure all appear to influence the level of heterogeneity of ruptures. Combining the results from each of the above analyses, we find that that rupture complexity is not limited to large magnitude events, and the factors that control rupture complexity for large magnitude earthquakes also control them for smaller events.

We derive a local earthquake catalog and detailed 3D models of seismic P- and S-wave speeds (vP and vS, respectively) for the region around the 2019 Mw7.1 Ridgecrest earthquake in the Eastern California Shear Zone. The study uses traveltimes from aftershocks of the Mw7.1 Ridgecrest earthquake recorded by seventy-eight geophone sensors deployed rapidly in two rectangular arrays (Catchings et al., 2020) and seventy-four broadband seismometers from the regional network. The raw waveform data are first processed using automated procedures to detect and locate nearly 95,000 earthquakes--more than two and a half times the number of events reported by the published catalog of the Southern California Seismic Network (36,343 events). Over 2.47 x 106 P- and 1.52 x 106 S-wave arrival times are used in a new traveltime tomography method based on stochastic Voronoi-cell subspace projections of the model space (Fang et al., 2019) and a robust 3D ray tracer that uses the Fast Marching Method for solving the eikonal equation (White et al., 2020). The derived high-resolution vP, vS, and vP/vS models correlate closely with surface geology and spatial seismicity patterns. Salient features include (a) a volume of low vP, vS, and vP/vS material at the SE terminus of the Eastern Little Lake Fault (ELLF) where it abuts the Garlock Fault; (b) high vP/vS values characterizing the crust hosting the majority of aftershocks on the ELLF; and (c) low vP, low vS, and high vP/vS beneath the Coso Volcanic Field.

During Wednesday's seminar, I will describe our earthquake catalog, velocity models, and key methods used to derive them. Then I will interpret the velocity models in relation to spatial seismicity patterns and surface geology. Finally, I will speculate on potential relationships between seismicity patterns, velocity structure, and rupture dynamics.

Paleoseismic rupture histories provide spatio-temporal models of earthquake moment release that lengthen the instrumental catalogue and can be used to evaluate numeric rupture models and forecasts. I review a rupture model designed to produce the fewest and thus largest magnitude earthquakes using paleoseismic dates from the southern San Andreas and San Jacinto fault (Scharer and Yule, 2020). I highlight new paleoseismic data from the San Gorgonio Pass region that restrict ruptures along the "ShakeOut" rupture path and suggests these ruptures alternate with periods of activity in the Cajon Pass region. Overall, the rupture history supports inferences from dynamic and quasi-static rupture modeling. Based solely on the ages of paleoearthquakes, the maximum rupture model produces rupture frequencies and a magnitude distribution similar to the historic record and a comparison of independent data of slip per earthquake with the average slip along the faults derived from the rupture model shows that missed events are unlikely. Thus, although it is an end-member model, it should provide a reasonable comparison for magnitude-frequency distributions derived from earthquake rupture forecasts. Sub-regional clustering suggested by the model deserves additional study into the underlying physical properties that would promote such behavior in southern California and the timescales over which it could occur.

The Salton Trough hosts the southernmost segment of the San Andreas fault that is overdue for a large magnitude earthquake rupture. To better prepare for such an event, we need improved velocity models in the area that characterize the geology better and thus produce more accurate seismic hazard estimates. Using a combination of local earthquakes and active source data from the 2011 Salton Seismic Imaging Project (SSIP), we develop a high-resolution local travel time velocity model in Coachella Valley down to 10 km below sea level. The model shows strong correlations with the surface geology at shallow depths and lateral velocity changes across surface traces of mapped faults and seismicity trends at depth. We map the 3D asymmetric basin of Coachella valley, high-velocity (>6.4 km/s) zones possibly related to Late Cretaceous mylonitic rocks or older inherited basement structures, and a potential unmapped fault zone in Lost Horse Valley. Comparisons between the model and regional community models developed by the Southern California Earthquake Center (SCEC) reveal significant differences in basin geometry and crustal heterogeneity. We smoothly embed our SSIP-derived models for Coachella and Imperial Valleys into the SCEC models to produce hybrid models. Low-frequency seismograms (6 -- 30 s) generated from the hybrid models better fit observed waveforms than the SCEC models alone. The hybrid models currently represent the best models in Salton Trough for earthquake hazard assessment and geologic studies.

The paradigm of plate tectonics is fundamental to our understanding of the Earth, yet the question of what makes the lithosphere "plate-like" remains unanswered. As Earth's outer thermal boundary layer, the lithosphere derives its high viscosity largely from its cold temperatures, relative to the warmer asthenosphere. However, the roles of partial melt and volatiles in further reducing asthenospheric viscosity are still debated. Seismic wave conversions generated by a velocity gradient within the lithosphere-asthenosphere transition, combined with surface wave tomography, provide key constraints on these questions.

Beneath broad regions of young continental lithosphere, such as the western U.S. and much of Alaska, surface wave constraints on lithospheric thickness are compatible with the depths of lithosphere-asthenosphere velocity gradients implied by converted body waves. However, typical steady-state conductive models consistent with continental heat flow produce temperature and velocity gradients that are too gradual in depth to produce observed converted and scattered body waves. The presence of an additional factor, such as partial melt at the base of the thermal lithosphere, is needed to sharpen lithosphere-asthenosphere velocity gradients in many young continental regions. Anatolia provides an example of an asthenosphere that is particularly warm and rich in partial melt. In contrast, beneath cratons, lithosphere-asthenosphere velocity gradients are often gradual enough in depth to be consistent with a purely thermal origin.

In the oceans, many observations of scattered and reflected body waves indicate velocity contrasts whose depths follow an age-dependent trend. Modeling of fundamental mode Rayleigh waves from the Pacific ocean indicates that cooling plate models with asymptotic plate thicknesses of 85-95 km provide the best overall fits to phase velocities at periods of 25 s to 250 s. These thermal models are broadly consistent with the depths of scattered and reflected body wave observations, and with oceanic heat flow data. However, the lithosphere-asthenosphere velocity gradients for 85-95 km asymptotic plate thicknesses are too gradual to generate observable Sp phases, both at ages less than 30 Ma and at ages of 80 Ma or more. To jointly explain Rayleigh wave, scattered and reflected body waves and heat flow data, oceanic lithosphere can be characterized as a thermal boundary layer with an asymptotic thickness of 85-95 km that also contains other features, such as zones of partial melt from hydrated or carbonated asthenosphere, that enhance the lithosphere-asthenosphere velocity gradient.

Sedimentary basins are depressions in the earth's surface that consist of alluvial deposits and sedimentary rocks, which are geologically younger and have slower seismic wave velocities than the underlying basement rock, and also decrease in thickness towards their margins. An important characteristic of the seismic response of these basins is that seismic waves are often trapped and reflected due to their interactions with the deep structure of sediments and the associated variations in material properties, per Snell's Law, resulting in the generation of propagating surface waves in addition to the accompanying propagation of 'body' waves. Such basin effects are known from first-principles to depend on a complex series of factors related to three-dimensional basin geometry (shape, depth, etc.) and aspects of the source-to-site path (i.e., the manner in which incident waves approach and enter the basin). These complexities are implicitly included in three-dimensional ground motion simulations, which can predict strong basin effects, and are incorporated in a relatively simple manner in sediment depth-dependent components of semi-empirical ground motion models. Not surprisingly these two modeling frameworks provide different outcomes.

We investigate the benefits of regionalizing basin response in ergodic ground motion models. Using southern California data, we consider: (1) how should basin and non-basin site conditions be classified?; (2) how does mean site response and its variability differ for basins compared to non-basin locations?; (3) what are the variations in basin response between different basin structures, and how can this be parameterized for predicting ground motion intensity measures?; and (4) what are the similarities and differences related to the prior three questions for simulated and empirical estimates of ground motion?

We present results of research targeted towards answering these questions using a combination of earthquake ground motion data analysis and targeted, physics-based simulations. The results show distinct levels of site response in sedimentary structures having different geometries and geologic origins.