ALBERT

Description: Seismic waveforms from the Eastern Tennessee Seismic Network are corrected to the nominal Wood–Anderson (WA) torsion seismometer to obtain a total of 11,905 maximum trace amplitudes from 690 events seen on 50 different horizontal components to determine a local magnitude scale for the Eastern Tennessee Seismic Zone (ETSZ). We use the following distance-correction function –log 10 ( A 0 )=0.538( r /17)–0.0002516( r –17)+2.0, in which A 0 is the maximum amplitude measured in millimeters and r is the hypocentral distance measured in kilometers; this better agrees with reported moment magnitudes for larger events in the ETSZ. Using the normal 100 km distance for M L normalization severely overestimates M L , and we therefore chose to adopt the 17 km normalization technique. The –log 10 ( A 0 ) is very flat at distances 〉200 km, suggesting unusually low distance attenuation at local and near-regional distances from the ETSZ. The WA response reported by the International Association of Seismology and Physics of the Earth’s Interior to the nominal response shows no significant difference in the distance-dependent factor for the log A 0 term, although the revised response consistently yields M L values that are 0.1 lower than those found using the nominal response. The b -values for the currently reported duration magnitude are lower than the b -values obtained using the newly calculated M L scale. The relationship between M L and M D can be expressed as M L =0.68093 M D +0.64603. The catalog of events used in this study is complete for M L 〉1.3.

Description: The spatial variation of seismic b -value is mapped for the New Madrid seismic zone using local earthquakes occurring between 1995 and 2013. A region of high b -value at about 1.8 is found in the northern part of the Reelfoot fault. By performing probability tests and following different procedures to obtain the result, we show the anomalously high b -value is not a processing artifact and is statistically significant. We attribute the b -value anomaly to creep behavior on the northern segment of the Reelfoot fault. Creep behavior is suggested by the recently discovered possible tremor and by the presence of quartz-rich rock as indicated by low V P / V S ratios detected in a local earthquake tomography study. Because quartz is a weak mineral, ductile, creeping behavior could be facilitated at depth resulting in generation of earthquakes in the shallower, brittle crust. To the south of the Reelfoot fault, the b -value is about 1.2. The character of seismic activity along the Reelfoot fault clearly changes from north to south indicating a change in the physical state of the fault zone.

Description: Design of large-aperture broadband arrays and array stacking of waveforms for receiver function studies critically depend on the coherence of waveforms across an array. The coherence of teleseismic P and S waves in the 0.05–1.6 Hz frequency band has been examined using high signal-to-noise teleseisms recorded by the USArray Transportable Array. Instrument-corrected, time-windowed, and rotated P and S waves were filtered in five, single-octave frequency bands and then correlated to determine coherence in each band. The normalized correlation coefficient is used as a measure of relative coherence and plotted as a function of interstation distance, which is used as a proxy for horizontal wavelength. Up to ~100,000 unique station correlation pairs can be found for vertical component P and transverse component S . Coherence of teleseismic P waves across the USArray is seen to be uniformly high in the frequency band between 0.05 and 0.4 Hz, with average correlation coefficients of 0.8 or greater for over 10 horizontal wavelengths. P -wave coherence degrades at higher frequency, although coherence is still higher than 0.6 for lower mantle propagation paths. S -wave coherence is relatively less robust but is greater than 0.8 for the 0.05–0.2 Hz frequency band, again for lower mantle propagation paths. Average coherence drops for both P and S waves for frequency bands greater than 0.2 Hz at station distances influenced by wave propagation through the upper mantle. This characteristic of wave propagation in the Earth can be used to design high-resolution broadband phased arrays for detailed studies of large earthquake sources and Earth structure and increases confidence in using vertical stacks of P waves as effective source functions in regional receiver function analysis.

Description: A phased array of 19 broadband seismometers was deployed from November 2009 to September 2011 in an effort to detect nonvolcanic tremor or tectonic tremor associated with the Reelfoot fault, northern Tennessee. An autodetection algorithm using broadband frequency–wavenumber analysis was used to search for the recurrence of signals first reported during an active source experiment in 2006. The original signals appeared as short duration, impulsive arrivals with a high phase velocity ranging from 3 to 25 km/s. We have identified thousands of similar signals on the 2-year long array data. Two distinct detection peaks are observed with event azimuths from the west and northeast. The detections are most similar to the events seen in 2006 and are inferred to come from very small ( M L ~–1) microearthquakes that occur in the shallow basement on faults adjacent to the Reelfoot fault. These include detections with coherent S -wave energy that reinforce the interpretation of very small local and regional events. Other signals detected show distinct changes in slowness and azimuth as a function of time. These events were interpreted as atmospheric acoustic sources. The high-frequency content and impulsive arrivals of the nonacoustic arrivals are not consistent with tectonic tremor as seen in other parts of the world but do indicate seismic activity in the crust near the Reelfoot thrust fault that was previously unknown.

Description: The horizontal-to-vertical (H/V) technique by Nakamura (1989) was applied to data from 30 new field stations and 64 other broadband temporary and permanent seismic stations within the Mississippi Embayment of the central United States to develop a 3D model of unconsolidated sediment shear-wave velocity structure. Using the Dart (1992) map of sediment thickness as a basis, two self-consistent models of average shear-wave velocity versus sediment thickness were developed by utilizing the theoretical linear relationship between the frequency of the H/V peak and shear-wave velocity. One model was based on the observation that the H/V peak period T p (s) versus sediment thickness h (m) was seen to be approximately linear with the relationship T p =0.003266 h +1.084. The second model was developed by considering peak frequency f p versus sediment thickness parameterized to follow ln f p =8.325 x 10 –7 h 2 –0.00232 h –0.01796. Overall, the models show low-average shear-wave velocity near the edge of the Mississippi Embayment with velocities increasing with increasing sediment thickness, consistent with increased sediment compaction. These models will be useful in studies of site resonance and amplification for earthquake-shaking hazards and for wave propagation computations for the region.

Description: The traditional approach to both earthquake and Global Positioning System (GPS) location problems in a homogeneous half-space produces a nonlinear relationship between a set of known positions, seismic stations or GPS satellites, and an unknown point, an earthquake hypocenter or GPS receiver. Linearization, followed by an iterative inversion, is typically used to solve both problems. Although sources and receivers are swapped in the earthquake and GPS location problems, the observation equation is the same for both, due to the principle of reciprocity. Consequently, the mechanical part of the solution of the equations is the same and single-step closed-form solutions for the GPS location problem, such as the Bancroft algorithm, can also be used to solve for earthquake hypocenters in a homogeneous half-space. This article applies the Bancroft algorithm to synthetic and real data for the Charlevoix seismic zone and compares the location of ~1200 events estimated with both the Bancroft algorithm and HYPOINVERSE. The Bancroft algorithm shows quantifiable improvements in accuracy compared with traditional methods. We also show how tools commonly used by the GPS community, such as the geometric dilution of precision, can be used to better estimate the precision of the results obtained by a seismic network.

Description: Three small-scale refraction and gradiometry experiments were performed to investigate whether off-the-shelf exploration geophones and seismographs can be used to perform meaningful gradiometry measurements. Relative calibration of the geophones was attempted through huddle tests and spectral measurements of ambient ground motions. The results show that relative gains cannot be completely characterized because of geophone/ground interaction. Numerical tests show that typically observed 4.5% gain errors introduce a standard deviation of 0.03 s/km and 1.97° about the correct input slowness and azimuth, respectively, for 100,000 realizations of synthetic array data. A standard linear refraction experiment was performed to investigate the slowness of P and Rayleigh waves from hammer sources to compare with measurements taken from two gradiometer designs. One design consists of four 6-instrument gradiometers in a linear array to investigate the spatial and temporal variation of horizontal slowness and propagation azimuth for sources close to the array as well as to test the location abilities of the entire gradiometer array. Its location estimate for a shot 10 m from the center of the array using the two closest cells was close to the actual source position with a 1.38-m error. The gradiometers were able to correctly determine the slowness values of the P and surface waves identified in the refraction profile. A second gradiometer experiment involves superimposed cells to explore precision in calculation of spatial gradients. Significant increases in the precision of the wave attributes occur when a larger number of averaged time-shifted waveforms are used as the reference-station displacement in place of a single-reference-station displacement waveform. The concept of center-station correlation is introduced to avoid spurious amplitude errors from drastically affecting the wave parameter estimates. We conclude that the off-the-shelf equipment can be used to construct small dense gradiometer arrays that can be used to infer wave attributes that are important for the interpretation of wavefields in exploration seismology experiments.

Description: The Mississippi Embayment overlies the New Madrid seismic zone in the central United States and consists of a thick succession of unconsolidated sediments that covers the Paleozoic basement rock. The high-velocity contrast at the basement–sediment interface and the near-vertical ray paths of teleseismic P -waves generate very large P to S conversions on the radial component and large P reverberations on the vertical component. This characteristic of P -wave propagation is used to study P - and S -wave resonance in the sedimentary layer. Horizontal-to-vertical (H/V) and vertical-to-horizontal (V/H) power spectral ratios are calculated for a time window around the teleseismic P -wave arrivals at broadband stations of the USArray Transportable Array, Northern Embayment Lithosphere Experiment, and the New Madrid Cooperative Seismic Network in the embayment. Using a map of sediment thickness, we developed models for average P - and S -wave velocity-versus-sediment thickness. The resulting shear-wave velocity model matches very well with the model obtained from an H/V ambient noise study in the region. The shear-wave velocity model shows low velocity near the edges of the embayment with average velocities increasing with increasing sediment thickness, consistent with increased sediment compaction. Fundamental resonance frequency for S waves ranges from 0.2 to 0.4 Hz for the embayment sediments. Electronic Supplement: Two figures showing the synthetic test results and a table that includes peak resonance frequency and average V P and V S .

Description: Detailed shear and compressional gradient velocity structure for the unconsolidated sediments of the Mississippi Embayment (ME) have been obtained through simultaneous inversion of radial and vertical teleseismic transfer functions for 60 broadband stations inside the ME. Transfer functions were calculated by deconvolving a vertical array beam from the radial and vertical displacement at each station inside the embayment. The vertical array beam was constructed by stacking all the vertical components of motions for stations on the bedrock and outside the embayment to reduce the effect of incoherent scattering, improving estimates of the source time function and stabilizing the deconvolution. A bounded variable least-squares inversion technique was applied to invert for velocity nodes on the top and bottom of the sediment layer. P -wave velocity changes from 1.0 km/s near the surface and increases with depth from 3.5 to 4 km/s in deeper parts. S -wave velocity varies from 0.3 km/s and increases to 1.6 km/s in deeper sections to the south. Electronic Supplement: Gradient velocity structure of the northern Mississippi Embayment sediments for 60 broadband stations.

Description: Standard P -wave receiver function analyses in polar environments can be difficult because reverberations in thick ice coverage often mask important P -to- S conversions from deeper subsurface structure and increase ambient noise levels, thereby significantly decreasing the signal-to-noise ratio of the data. In this study, we present an alternative approach to image the subsurface structure beneath ice sheets. We utilize downward continuation and wavefield decomposition of the P -wave response to obtain the up- and downgoing P and S wavefield potentials, which removes the effects of the ice sheet. The upgoing P wavefield, computed from decomposition of the waveform at a reference depth, is capable of indicating ice layer thickness. This simple step removes the necessity of modeling ice layer effects during iterative inversions and hastens the overall velocity analysis needed for downward continuation. The upgoing S wave is employed and modeled using standard inversion techniques as is done with receiver functions at the free surface using a least-squares approximation. To illustrate our proof of concept, data from several Antarctic networks are examined, and our results are compared with those from previous investigations using P - and S -wave receiver functions as well as body- and surface-wave tomographic analyses. We demonstrate how our approach satisfactorily removes the ice layer, thus creating a dataset that can be modeled for crustal and upper-mantle structure. Solution models indicate crustal thicknesses as well as average crustal and upper-mantle shear-wave velocities. Electronic Supplement: Figure of measured data, the vertical-component stack used in deconvolution, and the resultant vertical, radial, and tangential transfer functions.