Publications

Abstract not provided.

This report describes a proof-of-concept method of seismic source discrimination using seismic gradiometry and a common machine learning technique. The tests described here are purely numerical, using synthetic seismic data and well understood mathematical techniques. The primary innovation described here is the application of a richer seismic data set derived from seismic gradiometry. Seismic gradiometry is a method to estimate the time variable spatial gradient of the wavefield to compute various wavefield attributes such as slowness, dynamic strain, and rotational motions. With the addition of these wavefield attributes, we are afforded up to twenty "compo- nents" of time series data measured at a single point on, or in, the Earth. This is in direct contrast to conventional three-component seismic data collected at several locations using a seismic network. Using the gradiometrically-derived wavefield components directly in a single-layer neural network, I show that it is possible to discriminate between three common seismic source types (earthquakes, explosions, and opening fractures) for various noise conditions and gradiometry configurations.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

As a part of the series of Source Physics Experiments (SPE) conducted on the Nevada National Security Site in southern Nevada, we have developed a local-to-regional scale seismic velocity model of the site and surrounding area. Accurate earth models are critical for modeling sources like the SPE to investigate the role of earth structure on the propagation and scattering of seismic waves. We combine seismic body waves, surface waves, and gravity data in a joint inversion procedure to solve for the optimal 3D seismic compres-sional and shear-wave velocity structures and earthquake locations subject to model smoothness constraints. Earthquakes, which are relocated as part of the inversion, provide P-and S-body-wave absolute and differential travel times. Active source experiments in the region augment this dataset with P-body-wave absolute times and surface-wave dispersion data. Dense ground-based gravity observations and surface-wave dispersion derived from ambient noise in the region fill in many areas where body-wave data are sparse. In general, the top 1–2 km of the surface is relatively poorly sampled by the body waves alone. However, the addition of gravity and surface waves to the body-wave data-set greatly enhances structural resolvability in the near surface. We discuss the method-ology we developed for simultaneous inversion of these disparate data types and briefly describe results of the inversion in the context of previous work in the region.

We invert infrasound signals for an equivalent seismoacoustic source function using different atmospheric models to produce the necessary Green’s functions. The infrasound signals were produced by a series of underground chemical explosions as part of the Source Physics Experiment (SPE). In a previous study, we inverted the infrasound data using so-called predictive atmospheric models, which were based on historic, regional-scaled, publicly available weather observations interpolated onto a 3D grid. For the work presented here, we invert the same infrasound data, but using atmospheric models based on weather data collected in a time window that includes the approximate time of the explosion experiments, which we term postdictive models. We build two versions of the postdictive models for each SPE event: one that is based solely on the regional scaled observations, and one that is based on both regional scaled observations combined with on-site observations obtained by a weather sonde released at the time of the SPE. We then invert the observed data set three times, once for each atmospheric model type. We find that the estimated seismoacoustic source functions are relatively similar in waveform shape regardless of which atmospheric model that we used to construct the Green’s functions. However, we find that the amplitude of the estimated source functions is systematically dependent on the atmospheric model type: using the predictive atmospheric models to invert the data generally yields estimated source functions that are larger in amplitude than those estimated using the postdictive models.

Abstract not provided.

Abstract not provided.

We document azimuthally dependent seismic scattering at the Source Physics Experiment (SPE) using the large-N array. The large-N array recorded the seismic wavefield produced by the SPE-5 buried chemical explosion, which occurred in April 2016 at the Nevada National Security Site, U.S.A. By selecting a subset of vertical-component geophones from the large-N array, we formed 10 linear arrays, with different nominal source-receiver azimuths as well as six 2D arrays. For each linear array, we evaluate wavefield coherency as a function of frequency and interstation distance. For both the P arrival and post-P arrivals, the coherency is higher in the northeast propagation direction, which is consistent with the strike of the steeply dipping Boundary fault adjacent to the northwest side of the large-N array. Conventional array analysis using a suite of 2D arrays suggests that the presence of the fault may help explain the azimuthal dependence of the seismic-wave coherency for all wave types. This fault, which separates granite from alluvium, may be acting as a vertically oriented refractor and/or waveguide.

This work is a follow-on guide to running the Weather Research and Forecasting (WRF) model from Aur et al, (2018), Building and Running 1 DAAPS Models: IFRF Postdictions. This guide details running WRF in a nudged configuration, where the u and v wind components, temperature, and moisture within a specified spatial and temporal window, are adjusted towards the observations, radiosonde observations in this case, using WRF's observation nudging technique. The primary modification to this methodology from Aur et al. (2018), is the use of the OBSGRID program to generate the nudging files and the updates to the namelist.input file. These steps, combined with those outlined in Aur et al. (2018), will generate a nudged WRF hindcast (or postdiction) simulation.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

Abstract not provided.

We invert far-field infrasound data for the equivalent seismoacoustic timedomain moment tensor to assess the effects of variable atmospheric models and source phenomena. The infrasound data were produced by a series of underground chemical explosions that were conducted during the Source Physics Experiment (SPE), which was originally designed to study seismoacoustic signal phenomena. The first goal of this work is to investigate the sensitivity of the inversion to the variability of the estimated atmospheric model. The second goal is to determine the relative contribution of two presumed source mechanisms to the observed infrasonic wavefield. Rather than using actual atmospheric observations to estimate the necessary atmospheric Green’s functions, we build a series of atmospheric models that rely on publicly available, regional-scale atmospheric observations. The atmospheric observations are summarized and interpolated onto a 3D grid to produce a model of sound speed at the time of the experiment. For each of four SPE acoustic datasets that we invert, we produced a suite of three atmospheric models for each chemical explosion event, based on 10 yrs of meteorological data: an average model, which averages the atmospheric conditions for 10 yrs prior to each SPE event, as well as two extrema models. To parameterize the inversion, we assume that the source of infrasonic energy results from the linear combination of explosion-induced surface spall and linear seismic-to-elastic mode conversion at the Earth’s free surface. We find that the inversion yields relatively repeatable results for the estimated spall source. Conversely, the estimated isotropic explosion source is highly variable. This suggests that 1) the majority of the observed acoustic energy is produced by the spall and/or 2) our modeling of the elastic energy, and the subsequent conversion to acoustic energy, is too simplistic.

Abstract not provided.

Abstract not provided.

We invert far-field infrasound data for the equivalent seismoacoustic time-domain moment tensor to assess the effects of variable atmospheric models and source phenomena. The infrasound data were produced by a series of underground chemical explosions that were conducted during the Source Physics Experiment (SPE), which was originally designed to study seismoacoustic signal phenomena. The first goal is to investigate the sensitivity of the inversion to the variability of the estimated atmospheric model. The second goal is to determine the relative contribution of two presumed source mechanisms to the observed infrasonic wavefield. Rather than using actual atmospheric observations to estimate the necessary atmospheric Green’s functions, we build a series of atmospheric models that rely on publicly available, regional-scale atmospheric observations. The atmospheric observations are summarized and interpolated onto a 3D grid to produce a model of sound speed at the time of the experiment. For each of four SPE acoustic datasets that we invert, we produced a suite of three atmospheric models for each chemical explosion event, based on 10 yrs of meteorological data: an average model, which averages the atmospheric conditions for 10 yrs prior to each SPE event, as well as two extrema models. To parameterize the inversion, we assume that the source of infrasonic energy results from the linear combination of explosion-induced surface spall and linear seismic-to-elastic mode conversion at the Earth’s free surface. We find that the inversion yields relatively repeatable results for the estimated spall source. Conversely, the estimated isotropic explosion source is highly variable. This suggests that 1) the majority of the observed acoustic energy is produced by the spall and/or 2) our modeling of the elastic energy, and the subsequent conversion to acoustic energy, is too simplistic.