S11B-0164
Time-Dependent Source Moment Tensors of Chemical Explosions With Various Source Configurations
We have developed time-dependent source moment tensors for nine chemical explosions detonated in a dedicated source-characterization field experiment. These explosions have different sizes, geometries, depths of burial and containment conditions. Two of the explosions also failed an adjacent vertical free face and resulted in horizontal material cast. These moment tensors provide insight into the source characteristics of these explosions, particularly the mechanisms of shear-wave generation from explosive sources. We employed a frequency-domain linear-inversion technique to obtain the moment tensors for these explosions. A detailed velocity model derived from refraction and surface-wave data was used. We also used surface-wave data to infer the medium attenuation required for the inversion. Near-source (< 700 m) broadband data were used in the inversion. The source moment tensors of all explosions are dominated by the isotropic component regardless of different source configurations. The time histories of the partially contained and uncontained explosions are more oscillatory than those of the contained explosions, possibly indicating the involvement of the explosion-free-surface interaction. The explosions that involved material cast have more complicated moment-tensor time histories. In addition, their moment tensors show an enhanced low-frequency (1-6 Hz) signal in the off-diagonal components, probably resulted from the material cast. Although the amplitude of the off-diagonal components is only a small fraction of the amplitude of the isotropic component, synthetic-seismogram calculation shows that it can generate shear waves of much larger amplitudes. This illustrates that although the deviatoric component of the moment tensor for an explosive source may be small compared with the isotropic component, it can be an important shear-wave generator.
S11B-0165
Explosion Discrimination in Eastern Russia Based on Amplitude Ratios Obtained From Analog Records
Both earthquakes and explosions occur in much of Eastern Russia. In general, earthquakes are concentrated in broad areas related to the boundaries between the Eurasian, North American, Okhotsk, and Amur plates. On the other hand, industrial explosions are related to mining, and the constructions of roads, rails, and dams, much of which occurs in the same general area as do natural earthquakes. The published and unpublished Russian bulletins (such as those from the Yakutsk and Magadan regional networks) contain information about both earthquakes and explosions however very few events are explicitly marked as explosions in those bulletins. This results in the contamination of the earthquake catalog by explosions which can result in a misinterpretation of local tectonics, and an erroneous assessment of the natural seismic hazard in the region. As it is impractical to re-examine all analog records obtained in the past 40 years, a rapid method of discriminating between earthquakes and explosions based on the regional network bulletins is desirable. As a preliminary effort, 106 explosions and 194 earthquakes recorded at seismic stations which used SKM-3 seismometers between 1985 and 1989 were analyzed to see if a discriminant could be obtained. Information about amplitudes, arrival times, and location of earthquakes and explosions were taken from both published and unpublished bulletins of the Yakutsk and Magadan regional networks and from analog seismograms obtained in Russia. Different Sg/Pg amplitude ratios are compared from explosions and small earthquakes. The ratios calculated are: Sg (h)/Pg (z), Pg (z)/Sg (z), Sg (h)/Pg (h), and Sg (z)/Pg (h), where h is the horizontal component and z is the vertical component. All amplitude ratios are plotted against distance, and energy class of the seismic shock (K class) as calculated by station. We applied a distance correction to the ratios and found that most consistent discriminants for separating earthquakes from explosions are the amplitude ratios Pg(z)/Sg(h) and Pg(z)/Sg(z) as a function of K class. There is a tendency of the explosions to have higher values than earthquakes. More distinct separations can be found if the ratios are plotted for specific stations and specific ranges of distances. The clearest distinction occurs at distances from 100 to 300 km.
S11B-0166
Moment Tensor Inversions of Single-Fired Mining Explosions at a Copper Mine in Arizona
A portion of the Source Phenomenology Experiments (SPE) conducted 9 explosions at a copper mine in Arizona in 2003 with different yields, depths of burial and confinement conditions. We used a frequency-domain linear inversion technique to obtain time-dependent source moment tensors for these explosions. A detailed velocity model for the inversions was compiled from the analysis of a refraction survey conducted at the test bed; first arrivals of P-waves recorded at the stations around the mine and the inversion of Rg waves generated by these explosions. The Green's functions were calculated using the reflectivity method. The model was tested against observed data through forward waveform modeling prior to completing the inversions. The resulting moment tensors for all sources produce horizontal force couples, Mxx and Myy that are similar in amplitude and shape and an enhanced vertical force couple, Mzz. This source asymmetry is related to interaction of the explosion with the free surface and results in enhanced vertical asymmetry in the moment tensor for the shallowest shots, some of which cratered the free surface. For the three, fully-contained explosions (B4, B6 and B10) with no cratering, the maximal amplitudes of the moment tensors increase with the yield reflecting the coupling effect. Strong off-diagonal elements of the moment tensor were found for all shots with the Mxz component often the largest. This deviatoric component may reflect the fracturing of the granite in the source region. The three, twice-Normal burden shots (B3, B5 and B8) all cratered resulting in large volumes of fractured granite and enhanced Mxz components that were bigger than the horizontal force couples, Mxx and Myy.
S11B-0167
Predicting Lg Coda Using Synthetic Seismograms and Media With Stochastic Heterogeneity
Recent examinations of the characteristics of coda-derived Sn and Lg spectra for yield estimation have shown that the spectral peak of Nevada Test Site (NTS) explosion spectra is depth-of-burial dependent, and that this peak is shifted to higher frequencies for Lop Nor explosions at the same depths. To confidently use coda-based yield formulas, we need to understand and predict coda spectral shape variations with depth, source media, velocity structure, topography, and geological heterogeneity. We present results of a coda modeling study to predict Lg coda. During the initial stages of this research, we have acquired and parameterized a deterministic 6 deg. x 6 deg. velocity and attenuation model centered on the Nevada Test Site. Near-source data are used to constrain density and attenuation profiles for the upper five km. The upper crust velocity profiles are quilted into a background velocity profile at depths greater than five km. The model is parameterized for use in a modified version of the Generalized Fourier Method in two dimensions (GFM2D). We modify this model to include stochastic heterogeneities of varying correlation lengths within the crust. Correlation length, Hurst number and fractional velocity perturbation of the heterogeneities are used to construct different realizations of the random media. We use nuclear explosion and earthquake cluster waveform analysis, as well as well log and geological information to constrain the stochastic parameters for a path between the NTS and the seismic stations near Mina, Nevada. Using multiple runs, we quantify the effects of variations in the stochastic parameters, of heterogeneity location in the crust and attenuation on coda amplitude and spectral characteristics. We calibrate these parameters by matching synthetic earthquake Lg coda envelopes to coda envelopes of local earthquakes with well-defined moments and mechanisms. We generate explosion synthetics for these calibrated deterministic and stochastic models. Secondary effects, including a compensated linear vector dipole source, are superposed on the synthetics in order to adequately characterize the Lg generation. We use this technique to characterize the effects of depth of burial on the coda spectral shapes.
S11B-0168
Rg to Lg Scattering Observations and Modeling
Lg is important to explosion yield estimation and earthquake/explosion discrimination, but the source of explosion generated Lg is still an area of active investigation. We investigate the contribution of Rg scattering to Lg. Common spectral nulls in vertical component Rg and Lg have been interpreted as evidence that scattered Rg is the dominant source of Lg in some areas. The nulls are assumed to result from non-spherical components of the explosion source, modeled as a CLVD located above the explosion. We compare Rg with 3-component Sg and Lg spectra in different source areas. Wavenumber synthetics and nonlinear source calculations constrain the predicted source spectra of Rg and directly generated Lg. Modal scattering calculations place bounds on the contribution of Rg to Lg relative to pS, S*, and directly generated S-waves. Rg recorded east and west of the Quartz 3 Deep Seismic Sounding explosion have persistent spectral nulls, but at different frequencies. The azimuthal dependence of the source spectra suggests that it may not be simply related to a CLVD source. The spectral nulls of Sg, Lg, and Lg coda do not correspond to the Rg spectral nulls, so for this overburied source, the spectral observations do not indicate that Rg scattering is a dominant contributor to Lg. Preliminary comparisons of Rg with Lg spectra for events from the Semipalatinsk Test Site yield a similar result. We compare Rg at 20-100 km with Lg at 650 km for Balapan and Degelen explosions with known yield and source depth. The events range from 130 to 50 percent of theoretical containment depth, so relative contributions from a CLVD are expected to vary significantly. For studied previously NTS and Kazakh depth of burial data, the use of 3-components provides further insight into scattering between components. In a complementary analysis, to assess whether S-wave generation is affected by source depth or scaled depth, we have examined regional phase amplitudes of 13 Degelen explosions with known yields and source depths. Initial Pn, the entire P wavetrain, Sn, Lg, and Lg coda have similar log amplitude vs. log yield curves. The slope of those curves varies with frequency, from approximately 0.84 at 0.6 Hz to 0.65 at 6 Hz. We will complement these results with similar observations of Balapan explosion records.
S11B-0169
Seismic Observations of Westdahl volcano and Western Unimak Island Alaska: 1999-2005
Westdahl volcano is a large basaltic shield volcano on the western end of Unimak Island Alaska in the Aleutian Island Arc. The volcano is topped by three separate vents, Pogromni Volcano, Faris Peak, and Westdahl Peak. The volcano is frequently active with known eruptions from Westdahl Peak in 1964, 1978, and 1991-92 that produced large basaltic lava flows. InSAR measurements indicate that Westdahl Volcano has been inflating at a slowly declining rate since 1992 (Lu et al., 2003). The Alaska Volcano Observatory has operated a network of six short-period seismometers on Westdahl Peak since 1998. Complementing this network are similar networks centered on Shishaldin and Akutan Volcanoes. Since 1999 more than 300 earthquakes have been located within 20 km of Westdahl Volcano. A volcano specific velocity model was determined for the western half of Uminak Island by simultaneously inverting for the velocity model and hypocentral earthquake locations using the program VELEST. Earthquakes located with the new model reveal five clusters of hypocenters: (a) a shallow cluster beneath Westdahl Peak, that largely occurred during a 24-hour period on January 7, 2004, (b) a concentration of 68 earthquakes with hypocenters ranging in depth from zero to eight km beneath Faris Peak occurring continually since 1999, (c) a diffuse cluster of long-period events northwest of Westdahl and Faris Peaks, (d) a cluster of 12 earthquakes near Pinnacle Rock, 12 km southwest of Westdahl Peak in October 2003, and (e) a cluster of 43 hypocenters near Unimak Bight, 20 km east of Westdahl Peak, that occurred between January and April 2004. Focal mechanisms were derived for four earthquakes in the Faris Peak cluster and four additional earthquakes that locate off the volcanic edifice (the four mechanisms are in the Pinnacle Rock cluster, the Unimak Bight cluster, and 20 km southeast and 30 km northeast of the volcano). Focal mechanisms in the Faris Peak cluster showed normal faulting with nodal planes trending north-south to northwest-southeast. Mechanisms of the off-volcano earthquakes are generally characterized by normal faulting with nodal planes trending southwest-Northeast. These events are consistent with a stress field dominated by the Aleutian subduction zone. The Faris Peak mechanisms are not consistent with the presumed regional stress field and may reflect volcanic process. Lu et al., (2003) proposed the observed inflation of Westdahl Volcano resulted from a slowly pressurizing magma source at 6 km depth beneath Westdahl Peak. The observed seismicity is consistent with this model. Lu, Z., T. Masterlark, D. Dzurisin, and R. Rykhus, 2003, Magma supply dynamics at Westdahl volcano, Alaska, modeled from satellite radar interferometry, Alaska, J. Geophys. Res. 108, 2354, doi:10.1029/2002JB002311, 2003.
S11B-0170
A Broadband Seismological Investigation of the Cameroon Volcanic Line
Volcanic lines without any age progression cannot be explained by the conventional plume model, and their is no consensus about their origin. One family of models proposes that hot lines can be explained by added complexities in the plume framework. These propositions include lateral transport of plume material, recurrent volcanism due to relict plume material in the mantle along an old plume track, and multiple plumes. Other models include lithospheric tensional cracks or so called ''leaky frature'', and upwelling limb of small-scale convection cells. The Cameroon Volcanic Line (CVL), a 1500 km long line of Cenozoic volcanic constructs, is one of the most prominent volcanic lines lacking a discernable age progression. About one-half of the length of the CVL occurs on land, allowing extensive study without the expense of seafloor instrumentation. The PASSCAL Cameroon project is a 2-year passive-source broadband seismic experiment, consisting of an 8 station prototype deployment in 2005 and a 32 station deployment in 2006. Data recorded at the stations distributed throughout Cameroon will be used to examine the seismic structure of the crust and upper mantle beneath the CVL, and evaluate models for the origin of volcanic lines without age progression. Eight stations were installed in January 2005 and operated successfully throughout the year. Preliminary results of receiver-function analysis of the Moho, 410, and 660 km discontinuities for these stations will be presented.
S11B-0171
Toward the Systematic Counting of Small Volcanic Seismic Events
One of the primary challenges facing modern volcano observatories is mining small signals from massive data streams in a consistent fashion (particularly seismic.) One such parameter is the number of earthquakes detected by a single seismic station - so called helicorder counts because they were traditionally performed by eye using paper helicorder records. While this measure is simplistic compared to reviewed located events, it has the advantage of being straight forward to implement and robust to network changes. At several volcanoes in the Aleutian arc, helicorder counts provide one of the few consistent multi-decade monitoring tools. As part of an NSF-supported REU (research experience for undergraduates) summer internship, we have developed an automated and expanded approach to helicorder counts with the goal of replacing and systematizing the current process. The challenge is to efficiently weed out non-volcanic signals which are easy recognized by eye (communications noise, calibration pulses, telelseismic and regional earthquakes). To do this, we use a simple event detector and several filtered data representations to measure basic parameters of each detected event including the duration, the dominant frequency, the maximum amplitude, and the impulsiveness of the first arrival. These parameters can then be used to identify and discard typical non-volcanic sources including communications noise, regional earthquakes and calibration pulses. We demonstrate the technique on several types of recent volcanic behavior observed at the Alaska Volcano Observatory.
S11B-0172
Activity of Earthquakes with Similar Waveforms Around Ontake Volcano, Central Japan
Ontake volcano area is known for its very high seismicity. Swarm activity has been repeatedly observed since 1976. The volcano erupted in 1979 and a large earthquake with a magnitude of 6.8 occurred in 1984 in the southeastern part of the volcano. A dense seismic network has been constructed for monitoring in this area. Recent studies using leveling survey reveal uplift above the swarm area. Earthquakes with similar waveforms (similar earthquakes) have been observed mainly at geothermal areas and plate boundaries. It is possible to use them to detect temporal change of crustal structure due to large earthquakes or volcanic eruptions using repeating similar earthquakes. With a view to use them as a monitoring tool, we search for similar earthquakes in Ontake area and examine temporal change of observed seismic waveforms. We use ~3300 events which occurred in this area from January 1999 to August 2000. First we pick up event pairs of which hypocenters are manually determined within 4 km. Then we calculate cross-correlation coefficients at each station from the waveform pairs band-pass-filtered at 4 to 8 Hz. The calculation uses 15-second time window that contains both P- and S-waves. We define similar earthquake pairs as event pairs with cross-correlation coefficients larger than 0.9 at more than 8 stations. Such event pairs are distributed mainly in earthquake clusters located to the east and southeastern flank of the volcano. The similar earthquakes have magnitude ranging from 1 to 2 and focal depths of 1 to 7 km. Occurrence intervals of the most pairs are very short, between a few hours and a few days. A few pairs have intervals of several months. But repeating groups with nearly the same interval have not been found due probably to the short data period. We will further analyze using expanded data period and area.
S11B-0173
Evidence of Velocity Variations During the Recent Mt. Etna Eruptive Activity Detected by Temporal Seismic Tomography
After nearly 10 years without any major flank eruption, volcanic activity resumed at Mt. Etna on July 17, 2001, giving rise to the first of the two most striking flank eruptions on this volcano in recent times. Fifteen months after the end (August 09, 2001) of this eruptive episode, a new eruption started abruptly on October 26, 2002 with only a few hours of premonitory seismicity accompanying the opening of eruptive fissures along a bi-radial direction. Since the end of this last eruption (January 2003), a period of weak volcanic activity occurred. On September 7, 2004 a new eruption occurred along a WNW-ESE to NW-SE oriented fracture system at the base of the South East summit crater. Compared to the previous two flank eruptions, the 2004 eruption did not have any measurable short-period seismicity and deformation variations. Since 2001, Mt. Etna is well covered by the INGV-CT permanent network and some temporary networks. This provides a unique opportunity to investigate seismic velocity variations before, during and after the three most recent eruptions. Characterizing spatial and temporal variations in seismic velocity in detail will yield a better understanding of the complex plumbing system beneath Mt. Etna and the triggering mechanisms for each eruption. The conventional way to detect temporal velocity changes is to separately invert velocity models for each data set and then examine their differences. This may, however, cause some artifacts in the velocity changes due to different data quality and distribution. Here we present a true temporal seismic tomography algorithm by constraining velocity models for different periods through a temporal smoothing operator. This technique considers the fact that the main features of the velocity models for different periods are similar. The temporal seismic tomography algorithm is based on the double-difference tomography code tomoDD that uses both absolute and differential arrival times to simultaneously determine event locations and the velocity model. Our preliminary results show that both Vp and Vs decrease underneath the central crater at depths of 0-5 km from the 2001 eruption to the 2002-2003 eruption, which may be related to the higher gas content of the magma intruded before the 2002-2003 eruption.
S11B-0174
Validating Seismic Tomography Results for the 1992 Eruptions of Mount Spurr, Alaska
In previous studies of Mount Spurr, Alaska, seismic tomography has been used to locate hydrothermally altered zones and magmatic pathways during the 1992 eruptions. In this study we investigate the applicability of new tomography techniques to this problem. We use double-difference tomography (tomoDD) to determine P-wave velocity structure and improved earthquake locations using catalog travel time picks. The challenge for this and similar studies is the lack of dense station coverage in the mountainous terrain surrounding Mount Spurr. In 1992, Mount Spurr erupted three times from its Crater Peak vent. Seismicity was most active at shallow depths at the summit (-2 to 1 km depth) and at depths of 0-5 km and 10-20 km at the Crater Peak vent. Based on our relocated hypocenters, seismicity in these three regions form clusters reflecting structures within the volcano. At the summit, the hypocenters define structures that dip shallowly to the south. Focal mechanisms indicate that these are normal faults. The two clusters of seismicity beneath Crater Peak define two vertical columnar conduits associated with magma ascent during the eruptive sequence in 1992. Tomography reveals P-wave low-velocity anomalies as low as 4.5 km/s at depths of 1-4 km and 10-15 km beneath the Crater Peak vent. These anomalies are co-located with the two areas of seismicity. No velocity anomaly was observed in the summit region, which has not erupted for over 5,000 years. To investigate whether our tomography study is well resolved, we calculate travel times for a synthetic model using the 1992 station and hypocenter coverage and reinvert for the synthetic velocity model. TomoDD accurately recovers the 4.5 km/s P-wave velocity bodies modeled in our synthetic dataset beneath Crater Peak, indicating that this technique is robust for Mount Spurr despite its relatively sparse network.
S11B-0175
Localization of scattered waves revealed from active sources around the Iwate Volcano, northeastern Japan
I have investigated the characteristics of seismic wavefield from active sources around the Iwate Volcano, northeastern Japan, by using the data from a dense seismographic array. The active seismic survey was carried out in October 2000 under the National Project for the Prediction of Volcanic Eruptions to clarify a three-dimensional velocity structure beneath the volcano. The survey was extensive with nine chemical explosions observed by 330 temporary seismic stations around the volcano. I examined the time evolution of wavefield by measuring band-pass-filtered rms amplitudes of seismic waves, and then spatially interpolated them to display an amplitude map for each sliding time window with a length of 1 s. The general feature of wavefield is a circular spread of wavefront from the explosion and gradual decay of coda amplitudes with time. However, there remains significant seismic energy at areas including the explosions and the summit. This localization of seismic energy is particularly evident for the explosion located at the eastern flank of volcano and in the frequency ranges lower than 8 Hz. The plot of amplitude versus epicentral distance takes spindle shape, suggesting a contribution of strong scattering. By applying a diffusion model to the amplitude data, I estimated a mean-free-path of about 800 m. This value is longer than that of 200 m estimated at the Vesuvius volcano, Italy, probably reflecting lower volcanic activity of Iwate volcano. The area of energy localization roughly coincides with a low-velocity (LV) zone beneath an area covering the summit and the eastern flank of volcano. LV materials are interpreted as relatively younger volcanic edifices that reflect the evolutionary history of volcano. From the spatial distribution of P-wave amplitudes corrected for geometrical spreading, the LV zone is characterized by moderate attenuation. The above observations suggest an effective excitation and entrapment of seismic waves by small-scale heterogeneities with large velocity and/or high density embedded in a LV matrix.