Moment Tensor Inversion in an Optimized, Three-Dimensional Seismic Earth’s Model

Abstract:

The robustness of moment tensor inversion for source characterization relies on the accuracy of the seismic model, and the kind of method used to compute Green’s tensor in it. Many moment tensor inversions are based on Green’s functions calculated with approximate methods, and are built on 1-D models or ad hoc 3-D averages. Three-dimensional seismic models and fully numerical Green’s functions can now be derived thanks to advances in high-performance computing and seismic techniques (e.g., adjoint methods and finite-difference or spectral element algorithms), but some issues remain: taking the crust into account, or imaging it, still is a difficult task; and if done, the necessary fine spatial discretization for numerical seismic wave propagation within this highly heterogeneous part of our planet, leads to time-consuming simulations, even for long-period signals. An eventual remedy is the construction of a crust with smooth, effective (or equivalent) seismic properties, using some filtering inherited from the homogenization theory. We have developped a non-linear, stochastic inversion procedure to generate 3-D models adapted to numerical simulation of the full seismic wavefield. This probabilistic approach is based on the parametrization of models using an optimized basis of smooth functions constructed by principal component analysis of a homogenized reference model. It provides a reduced and optimized parameter space for the Bayesian inversion of an ensemble of 1-D seismic models. An appraisal step of the ensemble of models is added to regularize the 3-D model laterally. Using CUB as a starting block, our procedure allows for the determination of a 3-D, effective, seismic Earth model, and of full 3-D Green’s functions – procured at a relatively low numerical cost with spectral element simulations. As an illustration, we then invert for moment tensors associated to events occuring in WUS, and compare the results to those obtained using a more classical procedure.

Media has not been submited for this Presentation

Development of Integrated Accelerographs Using Mems Technology with Efficient Real-Time Data Transmission and Deployment of a Collaborative Seismic Network.

Abstract:

Having a dense network of high-resolution accelerometers based in “Feedback” systems in a reasonable amount of time, is crucial for national seismic networks such as the Spanish seismic network IGN. However, this requirement is in conflict with the tight budgets that would made this virtually unreachable. IGN has designed, developed and manufactured integrated and comprehensive accelerographs that allow near real-time transmission of seismic data, based on what is known as MEMS technology. This type of instrumentation enables compliance with the requirement to produce accurate ShakeMaps based on large amounts of observed data, and not deducted (deducted of very simple analytical attenuation laws, almost in its entirety). It would also be very important for the development of Technical Building Regulations in areas of seismic risk. All of this with prices at least ten times cheaper than high-resolution accelerographs. With the aim to demonstrate its reliability, it has been realize a testing process of these MEMS accelerometers in a vibrant table at CEDEX and its comparison with a high resolution commercial accelerometer Guralp CMG-5T, and even the recorded accelerograms in 2013 seismic crisis at Torreperogil village, in the province of Jaén. Currently, the IGN is deploying several devices on focused areas such as Alhama Fault at Murcia Region (southeast of the Iberian Peninsula) and Aran Valley in the Catalonian Pyrenees. This new network is a densification of the existing accelerograph network based on standard commercial accelerometers and through volunteer citizens finds the installation places.

Media has not been submited for this Presentation

Advanced Structural Health Monitoring System for U.S. Department of Veterans Affairs Hospital Buildings

Abstract:

In collaboration with the VA, the U.S. Geological Survey has developed structural health monitoring (SHM) software that utilizes vibration inputs to continually analyze and archive the response characteristics of a building in near real-time. The SHM software is built on the Earthworm (EW) system (Johnson et al., 1995), which is an open data processing platform that allows any continuous waveform data to be collected into ring buffers from a digitizer for further analyses (http://www.isti2.com/ew). The SHM software initially determines baselines for a suite of structural response parameters, and then continuously examines the response for changes in these parameters. The structural parameters monitored currently are inter-story drift ratios, shear-wave travel times throughout the building, and base-shear capacity-demand ratio. The SHM software is integrated with a web-enabled SHM data management framework to support aggregation, storage, and reporting of SHM data obtained and analyzed from instrumented hospital buildings to record strong shaking from earthquakes. By analyzing and characterizing the threshold values for building-specific engineering demand parameters, the SHM software can determine inspection priority to be low, moderate, high or very high and thus assist efforts in evaluating the safety and integrity of buildings in the aftermath of an earthquake. The SHM software is scalable—to support an arbitrary number of sensors, and it is extensible—to accommodate new data streams without the need to rewrite storage and display logic. The SHM software works on site or remotely. The software was validated using both ambient and low- and high-intensity shaking data inputs to a full-scale seven-story reinforced concrete building section tested on the UC San Diego shake table.

Media has not been submited for this Presentation

Real-Time Completeness of the USGS ComCat Earthquake Catalog and Implications for Operational Aftershock Forecasting

Abstract:

Aftershock forecasts often depend on the characteristics of the ongoing sequence up to the time of the forecast. The observed aftershocks may be used to fit sequence-specific parameters for clustering models and to seed simulations of future aftershocks. However, the catalog magnitude of completeness may be elevated immediately following a large mainshock, due to decreased detection of smaller events, complicating the use of the early aftershocks. Simple functions have been found to describe the time-dependent magnitude of completeness following global (Page et al., BSSA, 2016) and California (Helmstetter et al., BSSA, 2006) mainshocks. As a further complication, the real-time earthquake catalog typically has omissions and errors not present in the final catalog. This real-time catalog incompleteness must also be quantified to avoid underestimating the probability of future aftershocks in real-time forecasts. We study the real-time completeness of the USGS ComCat catalog using snapshots downloaded periodically following selected M≥6 global and M≥5 U.S. earthquakes. We find that real-time catalog snapshots during roughly the first month of an aftershock sequence have a higher magnitude of completeness than the final catalog, with a typical difference of 0.3-0.8 magnitude units. We also find that the time-dependent magnitude of completeness equations developed using the final catalogs (Page et al., BSSA, 2016; Helmstetter et al., BSSA, 2006) can describe the completeness of the real-time catalogs during the first month with adjusted parameter values. Accounting for the real-time catalog completeness brings the aftershock productivity estimated from the real-time catalogs closer to the productivity estimated from the final catalog, increasing the accuracy of aftershock forecasts based on the real-time catalogs.

Media has not been submited for this Presentation

The mblg 3.9 September 13, 2017, Earthquake on the Virginia-West Virginia Border: A Significant Shock in the Giles County Seismic Zone

Abstract:

On September 13, 2017 the USGS NEIC reported a duration magnitude MD 3.2 earthquake at 37.473N 80.703W, depth 18 km near Lindside, West Virginia, close to the Virginia-West Virginia border. The earthquake was felt primarily in Monroe, Mercer and Summers counties, West Virginia and in Giles, Montgomery, Pulaski and Bland counties, Virginia. The maximum intensity reported to the USGS Did You Feel It? program was IV MM. The earthquake occurred in an area of moderate seismicity known as the Giles County Seismic Zone (GCSZ). The largest shock in the GCSZ occurred in 1897 near Pearisburg, VA, with mblg magnitude estimated from the felt area at 5.8. We relocated the hypocenter of the September, 2017 earthquake using a locally specific velocity model, at 37.4775N, 80.7035W, depth 21 km. We estimated the mblg magnitude at 3.90 +/- 0.26 using 26 stations at regional distances, and determined a duration magnitude MD of 3.71 +/- 0.17, using 33 stations. The duration magnitude is based on a correlation between the log of short-period signal duration and mblg. We determined a focal mechanism using 27 P polarities, 12 SH polarities and 16 SH/P amplitude ratios. The nodal planes with least rms amplitude ratio error are: strike N91E, dip 69 deg., rake -22 deg.; auxiliary plane strike N189E, dip 69 deg., rake -158 deg. This event is notable because it is the largest shock in the GCSZ since May, 1974 (mblg 3.7). This recent shock, like many others in the GCSZ, shares characteristics with those in the Eastern Tennessee Seismic Zone (ETSZ), which is also in the Appalachian Valley and Ridge province. The 2017 GCSZ focal mechanism is mostly strike-slip with a small normal component, on steeply dipping nodal planes trending approximately N-S and E-W. This type of mechanism is dominant in the ETSZ. Also, in both areas, focal depths tend to be greater than 12 km, unlike shocks to the east in the Blue Ridge, Piedmont and Coastal Plain provinces which tend to occur at shallower depths.

Media has not been submited for this Presentation

The 1 May 2017 British Columbia Earthquake Doublet and Implications for Complexity near the Southern End of the Denali Fault System

Abstract:

On May 1, 2017, two M6+ earthquakes occurred in northwestern British Columbia separated by about 2 hours. Despite their close distance about 10 km, the two events have different focal mechanisms, with first earthquake featuring a thrust focal mechanism and the second strike slip. Both focal plane solutions are inconsistent with slip on the nearby southeastern Denali fault system. To resolve their ruptured fault planes, we invert for the point source parameters, and analyze rupture directivity via waveform modeling. The results indicate that the first event ruptured upwards along a steep SW-dipping fault (strike 153°/dip 61°/rake 113°) with rupture length about 8 km, and the second event ruptured to the ESE along a left-lateral fault (strike 292°/dip 36°/rake 54°) with rupture length about 6 km. We infer that the earthquake doublet is related to slip on the active Duke River fault, and the involved faults could be associated with the transpression caused by collision of the Yakutat block.

Media has not been submited for this Presentation