S11B-1013 0800h
Regional Moment Tensor Parameter Resolution Tests
With the establishment of dense broadband seismic networks in many regions worldwide, regional moment tensor (RMT) inversion is becoming a routine tool for source parameter determination of weak ($M \geq 4$) earthquakes. We invert complete three-component regional waveforms to systematically investigate the influence of station distribution, crustal model, inversion frequency band, and assumed depth on source parameter (magnitude, depth, moment tensor) resolution. Synthetic tests involve pure strike-slip and dip-slip sources similar to the real data, the $M_W = 6.6$ July 26, 2001 Skyros, Aegean Sea earthquake and two $M_W = 4.5$ and 5.0 aftershocks. We define seven networks with average event-station distances from 420 to 1500 km and azimuthal gaps between $95-350\deg$. The networks are based on Skyros events and mimic typical configurations in the Mediterranean. We invert for all networks, 4 crustal models, 6 frequency bands (15-40 s to 50-100 s) and 12 source depths (4 to 36 km). Solutions close to the correct (synthetic) or an independent reference (observed) are selected. We find a complex interplay of inversion parameters, event size and type on source parameter resolution. Generally, more strike-slip than dip-slip solutions are selected, but depth is better resolved for dip-slip solutions; magnitude is well constrained. For dip-slip events, selected solutions depend strongly on the model. For strike-slip events, we obtain solutions for all models and good azimuthal coverage for larger events. For small events with few traces above noise level only a combination of close stations, good coverage and short-period analysis provides selected solutions. Generally, the true structure is unknown, therefore the best possible coverage and close-by stations should be used for RMT analysis. Phase mismatch at short-periods and long paths requires a careful selection of frequency bands considering crustal model knowledge and event-station distances. Preliminary tests with path-dependent models derived from a recent regional 3-D model do not show a significant resolution improvement over simple average 1-D models.
S11B-1014 0800h
Inversion for Seismic Moment Tensors in Anisotropic Media Using Standard Isotropic Techniques
Although seismic anisotropy is commonly observed in regions of the Earth's crust and mantle, it is usually neglected during the inversion for source parameters of earthquakes. Here, we investigate how this simplification may influence the inversion for seismic moment tensors. Our approach is based on synthetic waveform modeling for moment-tensor sources in anisotropic media. The results are used as an input for the inversion routine. The forward modeling shows that anisotropy may cause apparent volumetric source characteristics. On the other hand, the true nature of tensile sources may be hidden by anisotropic effects. Varying the orientation of the anisotropic elastic tensor in the neighborhood of the source and in the surrounding medium greatly influences the calculated waveforms. Our examples indicate that the fault-plane orientation can be derived accurately if additional constraints are used in the inversion process (such as assuming a zero trace of the moment tensor). However, even if the fault plane is known, non-double-couple components are usually underestimated. The examples further suggest that, if anisotropy is restricted to the near-source region, moment tensors and anisotropic elastic constants may be derived simultaneously form the observed waveforms. The degree of anistropy determined by this method is within 3% of the true value, whereas the orientation of the elastic tensor differs by up to 10$^\circ$.
S11B-1015 0800h
Relative Earthquake Location Using Surface Waves
Earthquake locations are fundamental parameters necessary for the in situ study of earthquake physics and faulting and are needed to map and to quantify Earth's deformation. Within dense continental seismic networks, earthquakes can be accurately and precisely located. However, for many important tectonic environments, existing catalog data are neither accurate nor precise. In particular, for isolated continental and off-shore areas, earthquake locations are estimated primarily using distant observations, which limits our resolution of hypocentral parameters. Surface waves are generally viewed as a complicated component of the seismic wave field. One can think of a number of reasons why using surface-wave phase shifts in event location might be complicated - most notable are waveform depth dependence and azimuthal variations in phase from faulting geometry changes. However, these problems are smaller than intuition may suggest, and phase shifts can be measured with better precision than one might think. Further, the large and relatively slowly varying horizontal slowness of the Rayleigh waves can produce accurate relative event epicentroid (epicenter of the event centroid) locations for events with similar faulting parameters. Using longer period observations (25-75 seconds) allows a stable linkage of events 10's of km apart. I illustrate these ideas with several examples, and show the value of using surface-waves to estimate relative earthquake epicentroid locations in remote areas. Specifically, I use double-difference phase-shift measurements from intermediate-period Rayleigh waveforms to constrain the relative locations of earthquake epeicentroids. I apply the method to about 40 strike-slip earthquakes in the vicinity of the Panama Fracture Zone. The resulting relocations more closely mimic the plate-boundary geometry than the initial NEIC locations and the mean phase shift residual is just above one second. Distances between the initial and final locations in preliminary inversions ranged from about 10 to 50 km. The median distance shift was 23 km. Centroid time shifts ranged from 0 to 4 seconds. The method provides an important tool for investigating spatial seismicity patterns associated with large events and their aftershocks, slow-slip events and background seismicity, and other important seismic processes operating in complex tectonic environments.
S11B-1016 0800h
Detecting and locating global seismic events using a matched filter on very-long-period seismic data
We apply a series of different matched filters to very-long-period (50 to 500 s) global seismograms of the GSN network from 1975 to 2002 in an effort to detect and locate previously unknown seismic events. These filters are sensitive to longer periods than the techniques routinely used to locate earthquakes, and have the potential to locate events that radiate most of their energy into surface waves: slow earthquakes (such as those along oceanic transform faults), glacial basal slip events (Ekstr\"{o}m et al., 2003), and perhaps avalanches and volcanic events. Such techniques may also compensate for the difficulty in detecting "normal" earthquakes where gaps exist in the broadband networks (e.g., southern oceans). We build on previous results by Shearer (1994) by examining a more complete global set of stations and applying filters that are sensitive to waveform polarity as well as amplitude. We compare the event detection and location capabilities of the different methods and present a list of any previously undetected events. We will estimate magnitudes for these events and compare them with cataloged events from the same regions to better understand the source mechanisms responsible for these very-long-period signals.
S11B-1017 0800h
Regional Seismic Travel-Time Prediction, Uncertainty, and Location Improvement in Western Eurasia
We investigate our ability to improve regional travel-time prediction and seismic event location using an {\it a priori,} three-dimensional velocity model of Western Eurasia and North Africa: WENA1.0 [{\it Pasyanos et al.,} 2004]. Our objective is to improve the accuracy of seismic location estimates and calculate representative location uncertainty estimates. As we focus on the geographic region of Western Eurasia, the Middle East, and North Africa, we develop, test, and validate 3D model-based travel-time prediction models for 30 stations in the study region. Three principal results are presented. First, the 3D WENA1.0 velocity model improves travel-time prediction over the {\it iasp91} model, as measured by variance reduction, for regional {\it Pg, Pn}, and {\it P} phases recorded at the 30 stations. Second, a distance-dependent uncertainty model is developed and tested for the WENA1.0 model. Third, an end-to-end validation test based on 500 event relocations demonstrates improved location performance over the 1-dimensional {\it iasp91} model. Validation of the 3D model is based on a comparison of approximately 11,000 {\it Pg, Pn, }and {\it P} travel-time predictions and empirical observations from ground truth (GT) events. Ray coverage for the validation dataset is chosen to provide representative, regional-distance sampling across Eurasia and North Africa. The WENA1.0 model markedly improves travel-time predictions for most stations with an average variance reduction of 25% for all ray paths. We find that improvement is station dependent, with some stations benefiting greatly from WENA1.0 predictions (52% at APA, 33% at BKR, and 32% at NIL), some stations showing moderate improvement (12% at KEV, 14% at BOM, and 12% at TAM), some benefiting only slightly (6% at MOX, and 4% at SVE), and some are degraded (-6% at MLR and -18% at QUE). We further test WENA1.0 by comparing location accuracy with results obtained using the {\it iasp91} model. Again, relocation of these events is dependent on ray paths that evenly sample WENA1.0 and therefore provide an unbiased assessment of location performance. A statistically significant sample is achieved by generating 500 location realizations based on 5 events with location accuracy between 1 km and 5 km. Each realization is a randomly selected event with location determined by randomly selecting 5 stations from the available network. In 340 cases (68% of the instances), locations are improved, and average mislocation is reduced from 31 km to 26 km. Preliminary test of uncertainty estimates suggest that our uncertainty model produces location uncertainty ellipses that are representative of location accuracy. These results highlight the importance of accurate GT datasets in assessing regional travel-time models and demonstrate that an a priori 3D model can markedly improve our ability to locate small magnitude events in a regional monitoring context. {\it This work was performed under the auspices of the U.S. Department of Energy by the University of California Lawrence Livermore National Laboratory under contract No. W-7405-Eng-48, Contribution UCRL-CONF-206386. }
S11B-1018 0800h
M$_{s}$-m$_{b}$ Scaling for Earthquakes in Central Europe and Implications for Seismic Discrimination at Low Magnitudes
Teleseismic M$_{s}$-m$_{b}$ is generally regarded as one of the best seismic discriminants, and there is high interest in extending this classical discriminant to smaller magnitudes using regional data. Lg waves are well suited for monitoring small events since they are usually the largest signals on regional seismograms. Earthquake studies of Nuttli's m$_{b}$(Lg) have verified its transportability in continental areas of Europe, Asia, and North America, and a transportable M$_{s}$-m$_{b}$(Lg) relationship for earthquakes would be a significant advance for broad-area, nuclear test monitoring. The potential for M$_{s}$-m$_{b}$(Lg) discrimination might seem problematic, since in essence, it involves the comparison of two S-based measures, as oppose to the classical discriminant involving a P-based m$_{b}$. Nevertheless, several studies suggest that discrimination using m$_{b}$(Lg) might be better at low magnitudes due to differences in the scaling rates of m$_{b}$(Lg) and m$_{b}$(P) for underground explosions. In this study, we make use of frequency-wavenumber processing of Grafenberg Array data to extract Rayleigh waves for European earthquakes, and develop an M$_{s}$-m$_{b}$(Lg) relation extending to M$_{s}$ 2.5. This relation compares well with observations in other regions, including the western United States. We assess the discrimination potential with M$_{s}$-m$_{b}$ observations for explosions compiled from other studies, including the m$_{b}$(Pn) study of Bonner and co-workers. Earthquake relations for m$_{b}$(Lg) and m$_{b}$(Pn) scale similarly and practically overlie for 2.5 $<$ M$_{s}$ $<$ 5.5. On the other hand, explosion relations scale differently, where event populations remain parallel for m$_{b}$(Lg), but merge at small magnitudes for m$_{b}$(Pn). At the smallest magnitudes in this study, populations are separated by $\sim$0.5 magnitude units for both m$_{b}$(Lg) and m$_{b}$(Pn).
S11B-1019 0800h
Differential Travel-time Residual Analysis for Improved Site Correction and Catalog Quality, and Insights into Laterally Varying Earth Structure in China / East Asia
We are exploring the use of relative travel-time residuals to better constrain models of the earth for use in event location, and to usefully augment station correction information, using residual estimates from events which were previously thought to have little utility due to poor ground truth (GT) parameters. The residual differences also prove to be valuable in identifying poor arrival time picks. For an event recorded at two stations, the travel-time residual difference is insensitive to the event location for certain geometries. The residual difference represents variations in properties of the earth for paths to the two stations, which have not been addressed by the velocity model. Because of the relative insensitivity to event location, relative constraints on travel-time residuals may be helpful in developing relative station corrections for source regions having little in the way of high-quality (ground truth) sources, thus improving the overall confidence of derived correction surfaces. By combining all relative residual information and performing an inversion for consistent differential patterns, we derive a 2D function that maps laterally-varying residual slowness deviations from the background model, as well as to obtain site terms based on the residual differentials. We apply the technique to events from the combined global and zABCE catalog, relocated using IASP91. We have selected event / station geometries such that a 20 km change in source location results in less than 0.2 second change in the travel-time residual differences, based on synthetic sensitivity studies. For this study we restict our test to source-receiver distances of 2 - 20 degrees. Application of the inversion time terms reduces differential residuals by 50%. We observe regional slowness patterns similar to those obtained by other authors using single ray (Pn tomography) techniques. These show fast mantle velocities in central and western China, in particular under the Tarim, Jungar, Tsaidam and Sichuan basins, and slower velocities beneath eastern China, including the continental shelf and Shanxi graben regions. Site terms are negative throughout eastern China, reflecting thinner crust, and are positive for western China, especially Tibet. Results will be used to test improvement in location based on the collection of ground truth events in the area. We will also perform relocation of the entire catalog, using the differential results both to provide catalog pick quality control as well as station correction terms. This will provide an improved basis for finer analysis of seismicity, including high-precision relative relocation where event clustering is observed.
S11B-1020 0800h
Constraining Relative Source Locations with Coda Wave Interferometry: Theory and Application to Earthquake Doublets in the Hayward Fault, California
The relative location of seismic sources is of importance for the location of aftershocks on a fault, for the positioning of sources in repeat seismic surveys, and for monitoring induced seismicity. In this paper we show how the seismic coda can be used to infer a measure of the relative source location of two identical seismic sources from the correlation of the waveforms recorded at a single receiver. The theory is applicable to an explosive source in an acoustic or elastic medium, and for a point force or double couple in an elastic medium. For an explosive source the relative source location is constrained to be located on a sphere, while for a point force and a double couple the relative source location can be constrained to be located on an ellipsoid whose symmetry axis is determined by the point force or double couple. We validate the theory with synthetic seismograms, and apply the theory to earthquake doublets on the Hayward fault, California. The distance between events obtained from the coda waves agrees with the distance obtained from the double-difference method.
S11B-1021 0800h
Amplitude-duration and other discriminants for seismically recorded hydroacoustic phases
We expand to a considerably enlarged dataset the investigation of the amplitude-duration criterion D introduced by Talandier and Okal [2001] for the purpose of discriminating between earthquakes (D $<$ 0) and explosions (D $>$ 0) as sources of T phases recorded at teleseismic distances by seismic T-phase stations. In the case of earthquakes, we confirm that "hotspot earthquakes" taking place in intraplate volcanic edifices (Hawaii) can feature D $>$ 0 when located close to a steep conversion slope. By contrast, all subduction and transform events have D $<$ 0, as do genuine intraplate ("abyssal") earthquakes. For subduction events, we further show that D correlates well with parameters characterizing the slowness of the source, such as the energy-to-moment ratio, and the T-Phase-Energy-Flux-to-moment ratio. We tentatively reproduce, on an independent dataset, the observation by Dziak [2001] of preferential excitation of T waves by strike-slip events, at least at small magnitudes. We also show that the 1998 tsunamigenic PNG underwater landslide had a significantly deficient D with respect to the mainshock and main aftershocks. An enlarged dataset of 77 documented underwater explosions fails to produce a single negative D. In a pattern reminiscent of hotspot earthquakes, underground explosions can feature positive or negative D. We further discuss advanced discrimination criteria, such as the exponent of the power-law fall-off of the ground velocity spectrum of the T phase, found to be much higher (in absolute value) for earthquakes (including hotspot events) than for explosions, the goodness-of-fit of such laws (found to be high only for earthquakes), and the dispersive nature of the spectrum at the lowest SOFAR frequencies, found to be characteristic of explosions. The combination of the amplitude-duration discriminant D and of the advance criteria allows the correct identification of the nature of practically all sources in our dataset. We discuss the case of a presumed underwater landslide and of a volcanic explosion, both of which occurred near the shores of the Big Island of Hawaii in December 2002.
S11B-1022 0800h
Can Earthquake Location Be Improved in Western China?
Errors in P travel-time tables (and hence Earth models) can introduce systematic bias into epicentre estimates. Here station corrections estimated for underground explosions at the Chinese nuclear test site are used to relocate earthquakes around the site. This results in a significant reduction in the variance of the residuals (the differences between observed and predicted times) for earthquakes within 250\,km of the test site, confirming, as hoped, that corrections estimated from the explosion times are valid for earthquakes. The application of known station corrections is unsuccessful at reducing the error if variance in the station corrections across the network is low or if one or more recordings has low signal-to-noise ratio resulting in large measurement errors. The maximum effective distance of the corrections from the site is now under investigation. The results should show how rapidly changes in crust and upper-mantle structure affect travel-times at increasing distances from the test site.
S11B-1023 0800h
Reverse Modeling for Seismic Event Characterization
The localization of seismic events is of utmost importance in seismology. Current techniques rely on the fact, that the recorded event is detectable on most of the stations of a seismic array. Weak events, not visible in the individual seismogram of the array, are missed out. We present a new approach, were no picking of events in the seismograms of the recording array is required. The observed wavefield of the array is reversed in time and than considered as the boundary value for the reverse modelling using a pseudo spectral or finite difference approach. Assuming the correct velocity model, these reverse modeled wavefield focusses in the hypocenter of the seismic event. The origin time of the event is given by the time where maximum focussing is observed. The spatial extent of the focus resembles the resolution power of the recorded wavefield and the acquisition. This automatically provides the uncertainty of the localization with respect to the bandwith of the recorded data. The new method is particularly usefull for the upcoming large passive arrays since no picking is required. It has a great potential for localizing very weak events, not detectable in the individual seismogram, since the reverse modelling stacks the energy of all recorded traces and therefore enhances the signal to noise ratio. The method is demonstrated by 2-D and 3-D numerical case studies for a simple and a complex subsurface model which shows the potential of the new technique. Events with a S/N ratio smaller than 1 are considered as well as velocity errors and its influence on the localization. Velocity errors lead to unfocussed wavefields which may be used to update the velocity model similar to the migration velocity analysis in reflection seismology.
S11B-1024 0800h
Synthetic Earthquake Location Tests in Southern California
Our goal in this study is to improve earthquake locations in southern California by understanding the biasing effects of 3-D velocity structure on the observed travel times. We generate synthetic data sets using the actual station and earthquake locations and the pick distribution for data between 1984 and 2002. We begin by considering two test regions: Region A, which encloses much of the San Jacinto and Elsinore Faults and includes 48,050 events with 544,413 P-picks and 156,681 S-picks recorded by 185 stations; and Region B, centered on the Big Bear region, which includes 31,502 events with 517,981 P-picks and 106,957 S-picks recorded by 192 stations. Following single-event location based on optimal 1-D velocity models, we examine the residual patterns to gain insight regarding the scale of velocity variations and station terms. We present maps of source-specific residuals for different stations and plots of residual correlation and MAD (Median Absolute Deviation) differences between event pairs binned by their separation distance. We estimate random picking error for P and S picks from the intercept of the MAD plot as the event separation distance goes to zero. Next, we create synthetic data sets with the same source-receiver geometries as the real data and approximately the same statistical properties by computing travel time anomalies for random 3-D velocity structures. Our results provide insight regarding how to adjust parameters in the source-specific station term (SSST) method to match residual statistics and obtain optimal results. We will present location results for these regions, obtained use both using phase pick and waveform cross-correlation data.
S11B-1025 0800h
Relocation of a Cluster of Earthquakes in the Wabash Valley Seismic Zone
In 1995-1996, a regional network of 10 three-component seismometers and a 10 element three-component phased array was deployed in the Wabash River valley of southern Illinois and Indiana. Previous analysis of 83 days detected 232 earthquakes. Nearly 78% of these earthquakes were located near the town of New Harmony, IN and form what we refer to as the New Harmony cluster. We finished processing the remaining 128 days of data and 417 more earthquakes were added to the catalog. The majority of the coherent signals during this deployment were distinguished as quarry blasts, while at most 20% represent earthquakes. Beamforming from the array produced backazimuths and slowness vectors used for initial locations, however accuracy is limited by the low-amplitude, emergent nature of the signal. The geometry of this cluster currently extends 40 km southeast of the array. Since the earthquakes display similarities in phase arrivals, amplitude, and frequency content, we suggest a more closely spaced grouping. Ongoing work continues on a cross-correlation approach to improve initial locations.
S11B-1026 0800h
Comparison of Waveform Cross Correlation Performance Across Different Broad Area Seismic Source Regions
We are evaluating a method of locating seismic sources (earthquakes, explosions) based on the large-scale use of waveform cross-correlation (WCC) measurements instead of the conventional measurements of seismic wave arrival time (phase picks). WCC measurements have been demonstrated to be 10 to 100 times more accurate, where they can be obtained. The principal issue we are exploring is the extent to which a significant fraction of seismicity can be located using WCC measurements. We are studying the Charlevoix region in eastern Canada, the New Madrid seismic zone in the central United States, and northern California. In the first two regions the datasets have been assembled from scratch working in conjunction with regional network operators. For Charlevoix, courtesy of the Geological Survey of Canada, we now have 2,470 events with corresponding catalog, phase, and waveform data. For New Madrid, two datasets have been acquired. The first is from the PANDA deployment between 1989 and 1992, which consists of 884 events with bulletin and waveform information. The second is from the Center for Earthquake Research and Information (CERI) network, operated by University of Memphis, from which we currently have waveform data for 1995-2003, with extensive catalog and phase data. Preliminary WCC results of the PANDA network indicate that 68% (597 out of 884 events) correlate with cross-correlation coefficients (CC) above 0.7 at four or more stations. Four stations are the minimum required to obtain a location estimate. Both P- and S-waves are correlated on all three components. The window lengths are 1 s and the lags searched over are also 1 s. It appears in a few examples that similar correlations are possible over 1 to 10 kilometer inter-event separation distances due to a site resonance from soft sediments underneath certain stations. In the Charlevoix seismic zone there are only 10% of the events that meet the criteria of CC $>$ 0.7 at four or more stations. To explain this low fraction of similar seismicity we find that both earthquake density and station density are important for applying waveform cross correlation. A final comparison can be made with all of northern California where both earthquake density and station density are high. A recently completed project performing 26 billion correlation measurements for 225,000 events demonstrated that a remarkable 90% of the events in northern California have CC $>$ 0.7 with at least one other event at four or more stations. Subsequent work in California is using this data to relocate all the seismicity and look for repeating events which have high CC and differential times close to zero. Initial results indicate that repeating events appear across a variety of tectonic settings.
S11B-1027 0800h
Source Phenomenology Experiment in Arizona: Amplitude Decay of Regional Phases as a Function of Distance Between Hard and Soft Rock Mines
Monitoring for small nuclear explosions remains of critical importance to the nation, especially with the potential proliferation of nuclear weapons. However, monitoring for small explosions becomes complicated due to complex Earth structure, which may mask the signals generated by small explosions. Nuclear explosions can be simulated from chemical explosions, and this simulation can aid us in understanding the source phenomenology of small nuclear explosions. The first step we take to characterize the phenomenology is to study the amplitude decay of regional phases with respect to distance. A data set of single-fired, contained explosions and delay-fired explosions (production mining explosions) in both hard and soft rock mines were collected in Arizona over the summer of 2003. The varying geology of the two mines helps us determine a unique and robust criterion to distinguish the two sources. Since, similar signals are produced at each mine, but traveling in opposite directions we can compare the two different signatures and infer more accurate propagation parameters that identify between the two sources. The soft rock mine is situated in NE Arizona on the Colorado Plateau and is a coalmine. From the Colorado Plateau, the station profile follows a southern trend through the transition zone of the Colorado Plateau into Basin and Range tectonics, and has significant topography variations exist along the profile. The hard rock mine sits in the Basin and Range province of SE Arizona and produces copper. The station spacing was 3 km for the near source and extended to 40 km for the center of the station profile, with respect to both of the mines. All of the shots were varied in configuration. We will correct each shot's amplitude for station and site effects, using the coda amplification factor. The amplitude will then be calculated for both body and surface waves. We will then compare the amplitude decays with distance for all shots.