S31A-1868
Surface Wave Dispersion Measurements from Ambient Seismic Noise Analysis in Italy
We present the surface wave dispersion results of the application of the ambientnoise method to broadband data recorded at 114 stations from the INGV national network and the MedNet network. Vertical-component ambient noise data from October 2005 to March 2007 have been correlated with one another to compute estimated fundamental model Rayleigh wave Green's functions. Cross-correlations are calculated in one- hour segments, stacked over 1.5 yr, and Rayleigh wave group dispersion curves at periods from 7 to 45 s were determined using the multiple-filter analysis technique. The study region was divided into 0.2° × 0.2° to invert for group velocity distributions. Checkerboard tests were first carried out, and the lateral resolution was estimated to be in 0.6°. The resulting group velocity maps from 7 to 45 s showed good correlations with known geological and tectonic features, the signatures of sedimentary basins and crustal thickness are revealed clearly in the maps. The central Alps is clearly resolved with lower group velocities due to thick terrigenous sediments and deep Moho underneath the Alps. Our results also indicate that the Moho depth on the Tyrrhenian side of the northern Apennines is much shallower than the Adriatic side.
S31A-1869
Nationwide Ambient Noise Imaging of New Zealand: Constraining Crustal and Upper Mantle S-Velocities Using Surface Wave Dispersion Curves
Ambient noise tomography has been shown to yield good results in both continental and oceanic
environments. We extend Lin et al.'s pilot study of ambient noise tomography in New Zealand by increasing
the number of stations and extending the timespan of the dataset. In particular, we incorporate data acquired
during several temporary deployments including CNIPSE (34 stations; 6 months), SAPSE (30 stations; 14
months), NORD (5 stations, 8 months) and Marlborough (7
stations; 6 months) and data from newly operational permanent GeoNet stations (47 vs. 42 used by Lin et
al.).
We compute Rayleigh and Love wave dispersion curves, and then estimate lateral group velocity variations
and S-velocity versus depth profiles. The latter is done by using the neighbourhood algorithm, a direct
search method that enables us to estimate posterior uncertainties for model parameters from variations in
the dispersion curves.
In a test-case study, we have inverted two Rayleigh wave group velocity dispersion curves measured
between stations of the NORD deployment in the northernmost North Island. The results are consistent with
those of an earlier study by Horspool et al. of the same region using teleseismic receiver functions and
Rayleigh wave phase velocities.
http://www.geonet.org.nz
S31A-1870
Ambient Noise Correlations with Seismic Networks in Mexico
We present ambient noise correlations among several local and regional networks in Mexico. The results show strong arrivals in the primary and secondary microseism bands and also longer-period arrivals up 85 seconds. The correlations provide a method of joining detailed local studies into a larger regional map. The dense arrays, such as the MASE array in central Mexico often show coherent scattering that is evidence of major lateral variations in the crust. The travel times shows the response of the local fast anomaly associated with the Popocatepetl Volcano, near Mexico City. There is also a significant lateral change associated with the Trans-Mexican Volcanic Belt.
S31A-1871
The passive imaging of the Philippine Sea slab underneath Tokai region in Japan inferred from cross-correlation analysis of teleseismic S coda
The seismic interferometry has recently been developed practically and theoretically for monitoring subsurface structure. The spatial correlation of wave fields, such as noise and coda, at two receivers describes a wave field that is recorded at a receiver if there was a source at the other, the Greenfs function. If such a wave field between the two receivers can be retrieved, it would be possible to detect scattered or reflected phases from subsurface structure by taking migration with the cross correlated functions. In this study we applied the this technique to teleseismic S-coda observed at Hi-net tiltmeters, deployed by NIED, to investigate crustal and uppermost mantle structure, especially the Philippine Sea slab, underneath Tokai region in Japan. We used 246 teleseismic events with magnitudes 5.5 or larger for a period from April 2003 to Dec 2007, and applied bandpass filter of 0.05-0.5 Hz to S-coda on radial component. The time-window used in this study is 20 sec before and 500 sec after the S arrival. The specific features for calculating cross correlation functions used in this study is the followings that effectively enhance the effect of S-coda portion: we subtracted the effect of microseisms, depressed the portion of source-time function by applying taper, and removed the effects of later phases, such as ScS and SS. After these processing, we stacked the cross- correlated waveforms with good S/N ratios among all teleseismic events, and made cross-sections by taking migration technique. In the Tokai region, the Philippine Sea slab subducts west-northwestwards from the Nankai trough. A resultant migrated image for Tokai region clearly shows a dipping feature corresponding to the Philippine Sea slab, and this feature is good agreement with the hypocenter distribution. The dipping feature just below the hypocenters presumably corresponds to the oceanic Moho inside the slab, implying that this technique can be used to detect body waves. However, the continental Moho seems to be unclear in northern part of the cross-section, presumably due to a lack of station density. Moreover, another migrated image also shows a clear features corresponding to the Philippine Sea slab underneath central Japan.
S31A-1872
Ambient seismic noise study in Taiwan for two different scale arrays
It has been demonstrated that Time Domain Empirical Green's Function (TDEGF) from ambient seismic noise
cross-correlation can be used to investigate crustal velocity structure from many studies around the world.
For surface wave tomographic studies from ambient noise, the maximum exploring depth depends on the
aperture of receiver array and the lateral resolution relies on the density of station-pair paths. To decipher
subsurface structures in various scales, researchers can utilize some existing continuous-recording seismic
stations and/or deploy a newly dense receiver array in the study region. In this study, we perform
tomographic applications of ambient seismic noise analysis in Taiwan region for two arrays with very different
scales.
Taiwan is located at a complex convergent plate boundary zone where the Philippine Sea plate interacts with
the Eurasian plate. As a result, the lateral velocity variations show dramatic patterns among different geologic
provinces. In the past decade, many continuous-recording broadband stations have already been set up to
monitor earthquake activities in the Taiwan region. The BATS (Broadband Array in Taiwan for Seismology)
network is being operated by the Institute of Earth Sciences, Academia Sinica (IESAS) since 1994. Currently,
there are 20 permanent stations covering approximately 350 km by 400 km area around Taiwan, including
some remote islets. In this study we selected 7 years data (2000-2006) from BATS to get the TDEGFs which
were then used to measure inter-station phase velocities in the period band 5-30s. Finally we then
constructed 2D phase velocity maps. At shorter periods (5-10s), phase velocity distribution can compare well
with surface geology. At longer periods (14-22s), there is a saxophone shape low velocity zone beneath the
Taiwan Island.
Taipei Basin is a high-level artificial noise metropolis with a nearly triangular shape basin located close to
northern tip of Taiwan with area just around 20 km by 20 km, much smaller than the area BATS covers.
Central Geological Survey (CGS) entrusted IESAS to monitor seismicity in this region from 2004. There were
around 20 continuous-recording broadband stations with about 5km average inter-station distance. For this
study we selected 3 months data, from mid July to mid October in 2005, to calculate TDEGFs. Finally we
obtained 0.5-3s phase velocity maps, which can compare well with surface geologic structure. The days with
typhoon warnings were excluded from ambient seismic noise analysis due to the fact that TDEGFs are
affected by temporarily close and massive moving sources like typhoons. We also found that the source
direction of ambient seismic noise in typhoon days had close relationship with typhoon location.
S31A-1873
Crustal tomography of western Anatolia using ambient noise correlations
We investigate the crustal structure of Aegean area using continuous seismic records from the SIMBAAD temporary seismic experiment and permanent broadband stations in Greece and Turkey. This first study focuses on Western Anatolia where informations on the crustal structure is scarce. We use one-year noise records of a total of almost 120 stations with an average distance of 100 km. We compute daily noise cross- correlations and stack them for each couple of stations for all components. We apply a Multiple Filter Analysis (MFA) on reconstructed Rayleigh waves (0.02-0.1Hz bandpass) which gives us dispersion curves for all pairs of stations. Group velocity maps are estimated for different frequencies and inverted for S-wave velocity profiles. This study aims at constructing a reliable crustal model for mantle tomography. We also want to investigate the relation between crustal structure and surface tectonics at the transition between north-south extension in Western Anatolia and compression in Eastern Anatolia.
S31A-1874
A Slower and More Dangerous Seattle Basin - Application of Noise Correlation to Shallow Basin Surface Waves
Seismic hazard assessments depend heavily on near-surface S-wave velocity models. Here we present a basin-wide 3-D shear wave velocity model based on direct observations, rather than inferred from P-wave velocities. Shallow S-wave velocities are determined by near-surface geology, which in the Seattle area is a combination of crystalline rock, glacial till, young alluvium, and artificial fill. Because geology and therefore ground shaking can vary considerably within such an urban area, improving the accuracy and resolution of shallow S-wave models is key to improving seismic hazard assessments and predictions for ground shaking. We use 1-10s period Rayleigh waves extracted from the cross-correlation of ambient noise at station pairs within the SHIPS array, supplemented by nearby pairs of stations from Earthscope's Transportable Array and the Pacific Northwest Seismic Network, to solve for the 3-D shear wave velocity structure in the upper 4 km of the Seattle Basin. The SHIPS arrays consists of 80 3-component instruments recording continuously for two months with inter-station distances ranging from 2 to 90 km. We embed our shear-wave velocity model in a finite difference code to predict shaking in the Seattle Basin due to various earthquake scenarios. Our inversion shows that the shallow shear wave velocity is on average 20% slower than previous estimates, which will likely produce higher levels of ground shaking during an earthquake once we compute the finite- element simulations.
S31A-1875
Surface Wave Tomography of the Nechako Basin, British Columbia, Using Ambient Seismic Noise
The Nechako Basin in British Columbia, Canada has been a difficult basin to explore due to the presence of Tertiary volcanic outcrop. The volcanic outcrop makes the use of conventional seismic methods difficult due to a strong velocity inversion at its base. An alternative to active source methods is to examine the high- frequency surface wave field obtained from noise analysis, which is sensitive to large-scale crustal structure and has been successfully applied to measuring the depth of sedimentary basins. Ambient noise surface wave tomography will thus help to unravel the structural composition of the basin. We estimate station-to- station Green's functions within the Nechako Basin, by cross-correlating the seismic noise data recorded by 12 POLARIS and CNSN seismic stations between September 2006 and November 2007, using a two-station method. Rayleigh wave dispersion characteristics are then estimated from the vertical component of the seismic noise data by measuring within the microseismic band. Inversion of the dispersion curves is used to produce 2-D group velocity maps and 1-D S velocity models for Nechako Basin and the surrounding region. Preliminary results so far indicate that cross-correlation of the seismic noise data produced Rayleigh waves, and that the dispersion varies laterally across the basin. At the end of the project, we expect to be able to estimate the thickness of volcanics and sedimentary rocks, lateral variations, and crustal thickness within the Nechako Basin.
S31A-1876
Ambient Noise Interferometry and Love Wave Tomography in Southeast Tibet
In 2003-'04 MIT operated a seismograph network in SE Tibet and SW China to study crust and upper mantle structures and processes in the tectonically complex and seismically active region west of Sichuan Basin and around the Eastern Himalaya Syntaxis. Yao et al. (GJI, 2006, 2008) obtained high-resolution images of the crust and lithospheric mantle from surface wave array tomography with Rayleigh wave data from traditional two-station analysis and ambient noise interferometry. Using interferometry, an empirical Green's function (EGF) can be extracted from the time derivative of long-term cross-correlations for each station pair, Because of denser path coverage, a more uniform azimuth distribution, and higher frequencies, ambient noise tomography can resolve smaller scale structures in the shallow subsurface than traditional (ballistic) surface wave tomography. Here we report results of Love wave tomography. Compared to Rayleigh waves, EGFs for Love wave propagation have higher SNR at short periods (5-20s). The Love wave phase velocity maps inferred from EGFs are generally consistent with the Rayleigh wave maps. Using a model space search (Neighborhood Algorithm) we obtain from the phase velocity maps a 3-D model for shear wave speed (SH) with lateral resolution ~100km. The inferred VSH heterogeneity seems generally consistent with variations in VSV inferred from Rayleigh wave dispersion; conspicuous discrepancies are examined in view of non-uniqueness of the models and radial anisotropy. Both the Rayleigh and Love wave inversions reveal prominent low velocity layers in middle and lower crust. Such zones of low rigidity may influence -- or control -- crustal deformation in the region and constraining them better is important for better understanding of relative block motion and seismicity in this area.
S31A-1877
Mapping the radially anisotropic crustal velocity structure of NW Canada with ambient- noise tomography
We use ambient-noise cross-correlation to image crustal seismic-velocity structure in NW Canada. Our focus area surrounds the CANOE (CAnadian NOrthwest Experiment) array, a 16-month deployment of 59 broadband seismic stations. The geometry of the CANOE array was designed for studying the processes of continental accretion and the characteristics of continental lithosphere, and as such it extends from the Northern Cordillera on the west into the Archean Slave province to the east, crossing crustal terrains that span ~4 Ga of Earth history. We expand our study area westward and eastward by including 42 broadband stations from the Canadian National Seismograph Network and the POLARIS network. We estimate the Green's function for each pair of stations by cross-correlating day-long time series of ambient noise in the time period July 2004 -- June 2005. We observe fundamental-mode Rayleigh waves on cross- correlated vertical-component records and Love waves on the transverse components. All azimuths are well represented by the station coverage in this region, and the signal-to-noise ratios of the impulse responses are strongest for paths perpendicular to the Pacific coastline. We determine group velocities for the surface waves in the period range 5--30 s. Laterally, group velocities vary by as much as ±15% at the shortest periods and ±6% at longer periods, with the fastest velocities found within the Slave province and very slow velocities associated with thick sedimentary layers at short periods. We invert the group-velocity values (>2500 interstation paths) for 3-D radially anisotropic shear-wave velocity within the crust. The sensitivity kernels depend strongly on the assumed elastic structure, and we therefore use local kernels to account for the effects of laterally variable sedimentary structure. The model is further constrained by estimates of crustal thickness from receiver functions, LITHOPROBE reflection profiles, and CRUST2.0. We investigate whether the group-velocity data require anisotropic velocity structure or if they are equally well fit by isotropic velocity.
S31A-1878
Surface Wave Anisotropy in the Central Alps
We investigate the surface wave velocity structure of the central Alps based on phase velocity measurements made on broad-band, cross-correlated ambient noise. Continuous data from stations of the SDSNet and TomoCH arrays are filtered, cross correlated on a day-by-day basis, and then stacked. The correlation functions are processed with multiple filters and phase-matched filtering. Following filtering, phase information is unwrapped from the signals. The appropriate phase cycle is then chosen interactively, using a background dispersion model for comparison. Reliable phase velocity dispersion curves of 363 station- station paths for Rayleigh waves and 284 station-station paths for Love waves have been measured using this approach. Measurements are made between 5 and 100 second periods. The shortest periods have sensitivity even to the upper crust, whereas the longer periods sample well into the mantle. Using such broadband data greatly increases our ability to recover lithospheric structure. We invert these phase data with a linearized LSQR approach, solving for both isotropic and azimuthally anisotropic Rayleigh- and Love- wave velocity. Results are encouraging, and suggest we can resolve fast directions parallel to both the western Alpine oroclinal bend and the Rhine Graben. Furthermore, we image fast azimuths perpendicular to the orogen at longer periods and interpret this to be an imprint of subduction.
S31A-1879
Surface-wave Tomography of East African Rift System using Ambient Seismic Noise
The surface-wave tomography technique for the ambient seismic noise is applied to the east African rift system to investigate shallow crustal structures of the region. Even if the technique has been widely used in many regions to investigate crustal structure in the world, there have been difficulties in application of the technique to the east African region because of unstable data conditions of PASSCAL experiments. A meticulous check of record by record enables us of applying the technique to understand the tectonic environment of the region. The long-period data of one month showing good quality in cross-correlation results are used in this study. They are from the 1994-95 Tanzania Passive-Source Seismic Experiment for the Tanzania craton and its surrounding rift zone, and from the 2000-02 Ethiopia/Kenya Broadband Seismic Experiment and the adjacent permanent stations of the African Array for the Ethiopia rift. The Rayleigh- and Love-wave group-speed maps were inverted using LSQR algorithm for several period bands (5 - 50 s). The preliminary group-speed distribution maps yield results roughly consistent with regional geology. The tomographic images of the Tanzania region show a strong high velocity anomaly at the location corresponding to the Tanzania craton and low velocity anomalies at the surrounding rift regions. For the Ethiopia regions, the features of low velocity anomalies roughly agree with the Tertiary volcanic regions. Combining the Tanzania and Ethiopia broadband arrays, the outline of the east African rift system can be identified as the low velocity anomalies in the surface-wave tomographic results. The structural variation with depth and the feature of the regional shear-wave anisotropy of crust will be explored by converting group- speed dispersion curves into shear-wave velocity structure.
S31A-1880
The Lithosphere-Asthenosphere Boundary Beneath the Tanzanian Craton
The Tanzanian Craton, with dimensions of approximately 1000 km by 600 km, is centrally located within the East African Plateau between the eastern and western branches of the East African Rift. Being currently subjected to continental rifting, the Tanzanian craton is an excellent location for studying the interactions of an active rifting environment with a stable cratonic lithosphere. In this study we seek to find the depth of the lithosphere-asthenosphere boundary (LAB) beneath the Tanzanian craton using the S receiver function technique. The data are from the Tanzanian Broadband Seismic Network, which consisted of 20 broadband seismic stations deployed from May 1994 to June 1995. S receiver functions are computed using S phases from earthquake events at epicentral distances 60 to 85 degrees and SKS phases from 85 to 110 degrees. Individual S receiver functions are stacked by station. Only those receiver functions that show a clear Moho conversion phase and are consistent with each other are selected for stacking. Our preliminary results at a couple of stations show that the LAB is at ~65 km beneath the western rift zone and ~80 km at the Tanzania craton. From studying the lithospheric thickness beneath the Tanzanian craton, we hope to gain a greater understanding of deep continental structure as it evolves within a modern extensional regime.
S31A-1881
Ambient Seismic Noise Tomography in the Pacific Northwest Employing Earthscope Flexible Arrays
Ambient Seismic Noise is proving to be an invaluable tool for imaging the lithosphere and upper asthenosphere. Utilizing dense PASSCAL supported Flexible Arrays, permanent networks, and the Western extent of the Transportable Array, Ambient Noise Tomography (ANT) is being used to create a lithospheric model complementary to asthenospheric models in the Pacific Northwest. This model is created with denser station coverage than has previously been employed leading to increased resolution throughout the region. Previously documented large-scale structures are identified while new smaller scale heterogeneities are also investigated. The study region can be broken into two primary zones. The Cascadia Subduction Zone is instrumented through the FlexArray along Cascadia Experiment for Segmentation. This experiment is designed to better illuminate the segmentation of Episodic Tremor and Slip throughout the subduction zone, which has been linked to the crustal structure. The Mendocino Triple Junction links the Cascadia Subduction Zone with the San Andreas Fault system. The complex systems from surface to core mantle boundary around this triple junction are explored with the Flexible Array Mendocino Experiment. Employment of ANT in the Pacific Northwest is illuminating the complex processes and structures around the Mendocino Triple Junction and throughout the Cascadia Subduction Zone.
S31A-1882
Seismic Noise Tomography in the Chile Ridge Subduction Region, Northern Patagonia
We used cross-correlation of ambient seismic noise recorded in the Chile Triple Juntion (CTJ) region to estimate interstation surface wave time-domain Green's functions and to invert resulting travel times for crustal shear wave velocity. Data were recorded at the Chile Ridge Subduction Project temporary network of 39 broadband sensors deployed in the CTJ region by the Universidad de Chile, Santiago, and the University of Florida, from December 2004 to February 2007. Interstation distances ranged from 40 to around 100 km. We selected 30 days with low earthquake activity, and cross-correlated and stacked 24 hours of vertical component data at 34 stations pairs, resulting in nominally 1,122 travel-times along assumed-straight interstation paths. Velocities in two-dimensional cells of 20 km N-S by 30 km E-W extent were calculated using the method of Tarantola (1987). The process was applied to cross correlation pairs determined in three frequency bands, 0.1- 0.2 Hz, corresponding to shallow crustal velocities down to 15 km depth, 0.05 - 0.1 Hz, for velocities down to 30 km, and 0.01 - 0.05 Hz for deeper velocities. Our results show that cell velocities correlate well with known geologic features. We find high crustal velocities where the Patagonian Batholith outcrops or is likely present at depth, and low velocities correlate with the active volcanic arc of the Southern Volcanic Zone (e.g., Volcan Hudson, Volcan Cay). Slow velocities also correlate with well developed sedimentary basins and for the highest frequencies, the glaciers of the Northern Ice Field. A very high velocity anomaly in the southern Andes portion of the study area appears to correlate with outcropping metamorphic rocks and plutonic intrusions.
S31A-1883
Ambient Noise Rayleigh Wave Tomography in Antarctica From the TAMSEIS Array
We image the crust and uppermost mantle structure of the Transantarctic Mountains and surrounding regions of the Ross Sea and East Antarctica using Rayleigh wave group velocities obtained from the cross- correlation of ambient seismic noise. The data was recorded by the 43 broadband seismographs of the Transantarctic Mountain Seismic Experiment (TAMSEIS) which was the first large scale passive seismic deployment in Antarctica. The stations ran from December 2001 until December 2003, but due to the difficulty of running seismic stations in such extreme conditions we often obtain only 40-100 useful days of data for any station pair. To obtain travel times between station pairs we first normalize the recorded data with a running absolute mean to remove the effect of any earthquakes and then cross-correlate day long segments of the vertical component. Correlelograms from each station pair are then stacked, and then the positive and negative time lags are averaged and filtered at a range of periods between 5 and 29 seconds. Errors for travel times measured from the filtered correlelograms are estimated using a bootstrap technique. We obtain good quality waveforms with high signal to noise ratios for station pairs located on both rock and ice. Group velocity maps are obtained by inverting travel time at each period from all possible station pairs after visual inspection, signal-to-noise ratio, repeatability of measurement, minimum station distance, and number of days correlated have been considered. Each period is inverted separately. At long periods the maps show the fastest velocities in the Ross Sea region, where previous studies show the thinnest crust. At short periods, the fastest velocities lie over the Transantarctic Mountains and the slowest velocities lie over thickly sedimented regions in the Ross Sea. Slow velocities at short periods also appear over the Wilkes Subglacial Basin. A cross section of group velocities along a line of stations striking roughly North-South with 80 km interstation spacing also shows slow velocities in the vicinity of the Wilkes Basin indicating a previously unknown sedimentary basin . Further work will involve inversion of the group velocity maps for shear velocity structure and estimation of sediment thickness.
S31A-1884
Seismic Ambient Noise Tomography of Canada
We have applied ambient noise tomography to a comprehensive waveform dataset to study the surface wave dispersion characteristics of Canada. Our study area extends from 40°W (western Greenland) to 150°W (eastern Alaska) with 40°N as the southern boundary. Three-component broadband waveforms recorded over a 5-year period (2003-2007) at 519 stations, including EarthScope Transportable Array stations immediately south of the Canadian border, have been processed systematically. We follow standard procedures in preprocessing daily waveforms from individual stations, and then calculate the cross correlation function for every pair of stations to extract coherent surface wave energy propagating between them. A two-step stacking scheme is applied to the preliminary cross correlation waveforms to extract the Rayleigh wave dispersion curves from which group and phase maps at periods of 5-90 s are derived. Our results provide high-resolution images of many geological features, while the overall characteristics are in excellent agreement with global tomographic models published previously. In particular, the 8 s tomographic image reveals that the Canadian Cordillera is associated with a broad, prominent low-velocity zone extending both westward to the North American plate boundary and southward to the western United States. Eastern Canada, on the other hand, is underlain by a broad high-velocity zone except along the St. Lawrence River valley. Two additional efforts are planned for the near future, namely a similar study of Love waves and the incorporation of data from temporary deployments in places not spanned by permanent networks.
S31A-1885
Seismic Interferometry and the Spatial Auto-Correlation Method on the Regional Coda of the Non-Proliferation Experiment
A seismic recording of the non-proliferation experiment (NPE), made by a petroleum-exploration company in Railroad Valley (Nevada), contains the first break of the regional P phases followed by a three minute long coda. The transverse orientation and sign-bit recording of the array, renders distinguishing phase arrival times difficult. This motivates the use of seismic interferometry. The geometry does not permit recovering the Green's function from the first break arrival times using the stationary phase theorem, however the coda contains sufficiently equipartitioned energy. We study the result of seismic interferometry in the frequency-domain. This procedure is analogous to the spatial auto-correlation (SPAC) method, devised for studying microtremors by Aki in 1957. Cross-correlating two receiver stations retrieves, under favorable circumstances, an approximation of the Green's function between these two stations. To first order, this Green's function consists of a direct event traveling between the receivers. In the frequency domain, the lowest mode in the Green's function is a weighted and scaled zero-order Bessel function of the first kind, J0. We fit the frequency-domain of the recovered Green's functions to damped J0 functions. to recover phase velocity and estimates of the attenuation coefficients. Only energy between 1-4 Hz can be fitted unambiguously with J0 functions, because higher frequencies contain too much spurious energy. This result shows the equivalence of the SPAC method and seismic interferometry for the lowest mode in the Green's function. This study also shows how the coda of a regional event, seemingly unfavorably positioned, can contains energy useful for seismic interferometry.
S31A-1886
Direct Estimation of Love Wave Phase Velocities Using Circular-Array Records of Ambient Noise
We have derived simple and novel formulae to directly infer phase velocities of Love waves using two-
component, horizontal-motion, circular-array seismograms of ambient noise (microtremors). Our formulae
have been derived as an extension of the SPAC method (Aki, 1957), a popular technique of ambient noise
exploration using circular arrays. Most commonly, the SPAC method is used to infer phase velocities of
Rayleigh waves (and their dispersion curves) on the basis of vertical-motion data. When three-component
seismograms are available, it also provides the possibility to infer phase velocities of Love waves (and their
dispersion curves), a useful, additional constraint on the subsurface soil structure (Okada and Matsushima,
1989; Matsushima and Okada, 1990; Ferrazzini et al., 1991; Métaxian et al., 1997; Chouet et al., 1998;
Yamamoto, 2000; Saccorotti et al., 2003; Köhler et al., 2007). In the three-component SPAC method,
however, the Love-wave phase velocities have to be inferred only as part of the solution of a nonlinear
system of equations, in which the Rayleigh-wave phase velocities and the Rayleigh/Love power partition
ratios also appear as unknowns to be solved for. The Love-wave phase velocities are the only unknown to
appear in our formulae. All we need as the input is the two-component, horizontal-motion records of ambient
noise around a circle of radius r and at its center. We first define R(t,r,θ) and T(t,r,θ) as the
radial and tangential components, respectively, of the horizontal-motion seismogram at radius r and
azimuth θ as seen from the array center. The time series R1(t,r), T1(t,r) and T0(t,r) are
then defined as their weighted and non-weighted azimuthal averages:
R1(t,r)=∫-ππ
R(t,r,θ)exp(-iθ)dθ T1(t,r)=∫-ππ T(t,r,θ)exp(-iθ)dθ T0(t,r)=∫-ππ T(t,r,θ)dθ
Our new formulae state:
GR1T0(r,r;ω)/GR1T0(0,r;ω)=(J0+J2)(rω/cL(ω)) (SPAC+L method)
GT1T0(r,r;ω)/GT1T0(0,r;ω)=(J0-J2)(rω/cL(ω)) (SPAC-L method)
GT1T0(r,r;ω)/GR1T0(r,r;ω)=+i[(J0-J2)/(J0+J2)](rω/cL(ω)) (CCA-L
method).
On the left-hand side, GR1T0(r',r;ω) (where r'=r or 0) is the correlation function
between R1(t,r') and T0(t,r), while GT1T0(r',r;ω) is the correlation function between
T1(t,r') and T0(t,r) (ω denotes the frequency). On the right-hand side, J0 and J2 are
the zeroth- and second-order Bessel functions of the first kind, while cL(ω) denotes the phase
velocity of Love waves. Once the quantity on the left-hand side is known from measurement records, it is
quite straightforward to infer cL(ω) by simple inversion. We tentatively name the ambient noise
methods, based on our new formulae, as marked to the right of each equation above --- SPAC+L, SPAC-L
and CCA-L. Although earlier works have also presented elaborate circular-array ambient noise methods to
infer Love-wave phase velocities directly (Tada et al., 2006; García-Jerez et al., 2008), the ones we are
presenting here are by far simpler. We have tested the practical utility of our formulae by applying them to
both artificial and real ambient noise data. Preliminary results have suggested that the SPAC+L method has
the best performance, followed by CCA-L and SPAC-L.
S31A-1887
Microtremors Studies Using SPAC Method: Experiences and Applications in Mexico.
The study of Microtremors (environmental vibration) has become one of the key parts for the evaluation of seismic risk. That is because they can be used to estimate the site effects. The first approach is by obtaining the dominant period of the site through the H/V spectral ratio (Nakamura method). However there are parts of the body and surface waves that are not entirely known. The spectral H/V ratio is mainly influenced by SH resonance in the superficial layers. But if we are working with surface waves, like Rayleigh waves, they should be represented by the ellipticity coefficient of theoretical Rayleigh waves, whereas the absolute magnitude of the H/V spectral ratio can not be directly compared with the transfer function. The shear wave velocity of the shallow structure is a basic element in the studies of the ground amplification and for the site response of sedimentary basins. The SPAC method (Spatial Auto Correlations Method) was proposed by Aki in 1957, based on microtremors recorded in instrumental arrays. This method allows us to obtain the dispersion curve of the Rayleigh waves, from which we can estimate the velocity structure. In Mexico we have used this method in geotechnical applications, engineering and seismic hazard studies, to characterize the site effect. Since in some places we do not have enough information to validate our results, we use ellipticity curves to interpret the H/V spectral ratios and compare with our SPAC results. This was implemented in cities with substantial urban density as the city of Monterrey, Nuevo Leon and in areas where the site effect has had great impact as Mexico City and in areas with important seismicity as some parts of the state of Michoacan, There, it has been possible to estimate the shear wave velocity of the soil through this type of study.
S31A-1888
Continuous P-Wave Seismic Noise Observed At Regional Distance In California
We have observed continuous short-period P-wave seismic noise at regional distance in California. The data were recorded at two small aperture arrays about 400 km apart (Parkfield: 1 month of data, 11 km aperture; Mojave Desert: 6 months of data, 4 km aperture). Using time-domain normalization and stacking, and beamforming of the high-frequency (0.5-2.0 Hz) noise, each array reveals a continuous peak centered at a phase speed of about 5 km/s and at an azimuth of about 230 degrees. This coherent energy hiding in the background noise lasts for the duration of each data set. The energy shows no dispersion with frequency, but exhibits temporal variations of its frequency-slowness peak within the range of up to 0.1-0.25 s/km and 210-270 degrees. We explore possible explanations for these observations, including both cultural and natural sources of P-wave energy.
S31A-1889
Controlled-source seismic interferometry with one way wave fields
In Seismic Interferometry we generally cross-correlate registrations at two receiver locations and sum over an array of sources to retrieve a Green's function as if one of the receiver locations hosts a (virtual) source and the other receiver location hosts an actual receiver. One application of this concept is to redatum an area of surface sources to a downhole receiver location, without requiring information about the medium between the sources and receivers, thus providing an effective tool for imaging below complex overburden, which is also known as the Virtual Source method. We demonstrate how elastic wavefield decomposition can be effectively combined with controlled-source Seismic Interferometry to generate virtual sources in a downhole receiver array that radiate only down- or upgoing P- or S-waves with receivers sensing only down- or upgoing P- or S- waves. For this purpose we derive exact Green's matrix representations from a reciprocity theorem for decomposed wavefields. Required is the deployment of multi-component sources at the surface and multi- component receivers in a horizontal borehole. The theory is supported with a synthetic elastic model, where redatumed traces are compared with those of a directly modeled reflection response, generated by placing active sources at the virtual source locations and applying elastic wavefield decomposition on both source and receiver side.
S31A-1890
Source estimation and path corrections; noise vs earthquakes
The Cut-And-Paste(CAP) technique is applied to Wells earthquake in Nevada and Chino Hills earthquake in South California to study their mechanisms and depths, and produces surface waves travel time delays with respect to 1D models. Ambient Seismic Noise(ASN) cross-correlation method is then applied to station pairs with one of them close to the earthquakes, to get the time delays with respect to the same 1D model used in CAP. We compare these two sets of time delays with each other, and with predictions from existing tomography type models using a population of past earthquakes(Tan, et al, 2006) or ASN. The travel times agree within one second for those paths whose cross-correlations are better than 0.80. Thus, it appears that these two methods can be combined or used in stages to first determine the depth and mechanism (CAP) followed by surface wave delays (ASN) to fix the location; in short, locate events with ASN.
S31A-1891 INVITED
C3 (Correlation of Coda of Correlations): Improving the reconstruction of Green functions between stations of a network from noise records.
Our aim is to extract the Green function between every pair of receivers of a network, by recording only seismic ambient noise. A first method consist in correlating the noise recorded between each pair of stations. If the noise sources are evenly distributed, the noise correlation is the exact Green function including all types of waves. However in practice the noise is not isotropic. For this reason correlating noise provides only the Green's function accurately between a limited number of station pairs. Here we explore a new method, that consists in correlating coda waves reconstructed by noise correlations. This allows us to remove the effects of the directivity of the noise. We consider two stations A and B for which the Rayleigh waves could not be discerned in the correlation of continuous records of ambient noise. We computed all correlations between the station A (resp. B) and all other stations of the network. We select time windows in those virtual seismograms that correspond to coda and compute correlations between them. This meta-correlation is found to exhibit the surface wave part of the Green function between A and B that was not visible in the raw correlation of ambient noise. Further iterations are possible, showing that the C3 function contains also scattered waves. C5 functions clearly exhibits the direct Rayleigh waves. This procedure can be used to assess seismic velocity between stations, even in presence of a directive and poorly oriented ambient noise. It also justifies the use of the late part of the noise correlation functions for the monitoring of temporal changes of seismic velocities.
S31A-1892
Determining Secondary Microseism Source Locations With a Seismic Network
We developed a location method to determine the secondary microseism energy (5-8 seconds) distribution where the anticipated source region is in the near offshore and where there is a well-distributed onshore seismic network (Tian & Clayton, 2007). The procedure is adapted from a time-time earthquake location method (Baker et al., 2005). The method is based on a mesh of potential source locations. For each time window and each grid point we predict the response of a Rayleigh wave at every station in the seismic array, and then correlate the synthetic seismograms with the real data that is filtered to the microseism band. The zero-lag of the cross-correlation is then composited at the mesh point. We tested the location algorithm with synthetic seismograms due to a spatially distributed set of sources. It turns out that the locations of the microseism sources can be mostly well resolved; sources off the presumed mesh are not artificially mapped onto the mesh; and the image itself is robust up velocity errors of 10%. We then calibrated for the effect of a heterogeneous station distribution and the amplitude decay by systematically computing a test response for each point of the mesh and remove the gain. For an 06/08/2006 tropical storm, for example, the image is composed of many distributed patches, while the patterns change very rapidly with time.
S31A-1893
Studying the origin of deep ocean microseisms using teleseismic body waves
Recent studies of oceanic microseisms concentrated on fundamental-mode surface waves. One of the reasons for this is that extraction of fundamental-mode Rayleigh and Love wave Green functions from station-station correlations of ambient seismic noise has recently been demonstrated to be a very powerful tool for imaging of the Earth's crust and uppermost mantle. Extracting body wave Green functions from noise cross-correlations remains however problematic. In this study we concentrate on energetic arrivals in the microseismic frequency band that appear at near zero times in noise cross-correlations. We demonstrate their clearly seasonal behavior. For stations located in the Northern hemisphere, relative amplitudes of these arrivals are significantly stronger during the summer. Polarization analysis of cross-correlations of continuous records by stations of the Eastern Turkey Seismic Experiment (ETSE) helps us to identify this near-zero time signal as upcoming P waves. We determine the origin of these signals by simultaneously analyzing records at stations of seismic arrays located in Turkey, Yellowstone, and Kyrgyzstan. We define apparent slownesses of the studied arrivals beneath three arrays by applying seismic beamforming to inter-station cross-correlations and locate the regions where the signals were generated by using the ray tracing in a spherically symmetric Earth model. Our results show that the energetic arrivals seen at near-zero times in seismic noise cross-correlations are formed by teleseismic P, PP, and PKP waves. Similar to noise-forming surface waves, generation of this ambient body waves is strongly seasonal with sources located in southern and northern oceans during the summer and the winter, respectively. Moreover, body wave array analysis is accurate enough to unambiguously demonstrate that significant amount of the microseism's energy is generated far from the coast in deep oceans.
S31A-1894
Global surface wave tomography using seismic hum
Recently, Shapiro et al. [2005] obtained group velocity maps of Rayleigh waves at around 0.1 Hz in Southern California by cross-correlation analysis of long sequences of ambient seismic noise. This method is called ambient noise surface wave tomography. After the study, group-velocity maps have been obtained at local and regional scales but not at global scale. The global surface wave tomography requires low frequency data below 20 mHz. In the frequency range, background Rayleigh waves, known as seismic hums, have now been firmly established. Statistical examination of them shows that these waves must be excited randomly and persistently by globally distributed sources. This feature suggests that we can apply the technique of ambient noise tomography for data set of seismic hums. In this study we estimated phase velocity anomalies using data of seismic hums, and we inverted them for 3-D S wave velocity structure in the upper mantle. We analyzed 10-sec continuous sampling records from 1988 to 2000 at quiet 54 FDSN (Federation of Digital seismic networks) stations. We calculated cross-correlation functions between every pair of different stations on seismically quiet days [Nishida and Fukao, 2007]. We measured phase difference between the observed cross-correlation functions and synthetic ones [Nishida and Fukao, 2007] for 906 R1 paths and 777 R2 paths. We inverted the observed phase velocity anomalies for phase velocity maps from 3 to 10 mHz. Then, we inverted obtaining phase velocity maps for 3-D S-wave velocity structure in the upper mantle [Montagner, 1986]. At the depths shallower than 200 km, our results show low velocity anomalies associated with mid- ocean ridges and back arcs, and fast velocity anomalies in continental shield and platform area. Below 200 km our results show high velocity anomalies in subduction zones. These features are consistent with past studies.
S31A-1895
Recovering the Attenuation of Surface Waves From Noise Correlation : Synthetic Tests in a Spherically Symmetric Earth.
Cross-correlation of ambient seismic noise recorded by a pair of stations is now commonly recognized to contain the Green's function between the stations. Numerous works have used this property to measure the travel time of surface waves on inter-station paths. In the present study, we investigate the possibility of using noise correlation to measure the attenuation of surface waves. We carry out numerical experiments to generate seismic noise in an attenuating spherically symmetric Earth. Records of the wavefield produced by such noise are correlated, and obtained waveforms are compared with exact Green's functions of the medium. We show that the recovery of the attenuation depends on the distribution of the noise sources and the process applied to the noise records prior to the correlation (such as 1-bit normalization and whitening).
S31A-1896
Reflection Profiles Extracted From Ambient-Noise Using Seismic Interferometry
Seismic Interferometry (SI) is the process of generating seismic traces from the crosscorrelation of existing traces. One application of SI is the retrieval of surface-wave arrivals between two passive stations at the Earth's surface from the crosscorrelation of ambient noise. Another application is the retrieval of body-wave reflections from the crosscorrelation of ambient noise recorded at the Earth's surface. Retrieved reflections would afford the construction of subsurface velocity models and subsurface reflection images with higher resolution than provided by surface-wave tomography. So far the extraction of body-wave reflections has proven to be more challenging. Several factors contribute to this difficulty: e.g., the difference in geometrical spreading between body and surface waves and the reliance on a random distribution of noise sources in the subsurface, as opposed to the ubiquitous and well-studied surface noise. We apply SI to ambient noise and further process the retrieved records to bring out reflections. Approximately 11 hours of noise were recorded in a desert in North Africa on 8 parallel lines with 50 m station spacing and 500 m spacing between the lines. Strong surface-wave energy, concentrated mainly below 6 Hz, was caused by traffic along a road bisecting the survey in the Northern section of the survey. We therefore first applied a low-cut frequency filter, followed by a frequency-wavenumber filter to remove remaining surface-wave noise. The corner frequency on the high end was 24 Hz. Next, the traces were energy normalized and then crosscorrelated. Despite the relatively short recording period, we retrieve coherent events. A comparison of virtual common- shot gathers (a response from one virtual shot recorded by all receivers on a line) with common-shot gathers from an active survey along the same line, shows that the retrieved events coincide with reflections in the active data. We further process virtual common-shot gathers using routines commonly applied in exploration seismics to enhance reflections. The results are stacked sections, outlining subsurface structure along several lines. The Southern part of these stacked sections show several coherent events. In the Northern part these events are not traceable, probably due to remnant surface-wave energy. Comparing stacked sections retrieved using SI with the stacked sections from the active survey, we observe that two relatively shallow marker events in particular have been adequately reconstructed from the noise.
S31A-1897
Analysis of Bias in Surface Wave Phase Velocities from Ambient Noise Interferometry and an Iterative Approach for Azimuthal Anisotropy
Green's functions of surface wave propagation between two receivers can be estimated from the cross correlation of ambient noise under the assumption of diffuse wavefields or energy equipartitioning. Such interferometric reconstructions of the Green's function are generally incomplete, however, because the distribution of noise sources is neither isotropic nor stationary and the wave fields considered in the cross- correlation are generally non-diffuse. Furthermore, the empirical Green's function (EGF) is influenced by (unknown) properties of the medium between the receivers, making the problem non-linear. To prevent bias in ambient noise tomography we need to understand and account for incomplete Green's function reconstruction. With a far-field (plane-wave) approximation we analyze the effect of uneven ambient noise distribution and medium heterogeneity and azimuthal anisotropy on phase velocities measured from EGFs. The bias in measurement due to uneven noise distribution is generally larger than that due to medium heterogeneity and anisotropy and can be corrected if the noise distribution is known. The (normalized) azimuthal distribution of ambient noise energy can be estimated directly from the cross correlation functions obtained through ambient noise interferometry. The (smaller, second order) bias due to non-linearity can be reduced iteratively, for instance by using the tomographic model that results from the inversion of uncorrected data. Our method for estimating and reducing phase velocity bias due to uneven noise distribution is appropriate for seismic arrays with good spatial and azimuthal (interstation) path coverage and sufficiently far away from the dominant ambient noise sources (e.g., oceans). The correction of (azimuth dependent) phase velocity bias due to azimuthal variations in ambient noise energy is particularly important for the investigation of azimuthal anisotropy from ambient noise tomography. We illustrate the iterative procedure for bias suppression with data from a seismic array (26 stations) in SE Tibet, and we present new analysis on surface wave (azimuthal) anisotropy from ambient noise.
S31A-1898
The virtual refraction: useful spurious energy in seismic interferometry
Limitations in the source energy distribution in seismic
interferometry lead to spurious energy in estimates of the Green's function between receivers. Instead of
attempting to suppress spurious waves, we use one such spurious wave -- we call it the virtual refraction -- to
infer subsurface parameters. Numerical and real data examples illustrate how the slope of the virtual
refraction defines the velocity of the faster medium and the stationary phase point in the correlation gather
provides an estimate of the critical offset. We illustrate the conditions for improvement in signal-to-noise of
the virtual refraction, compared to conventional refraction analysis. This is due to the stacking of sources
inherent in the method of seismic interferometry. Just as conventional refraction analysis, applications of this
virtual refraction range from the near-surface (for example, mapping depth to the water table), to exploration
seismology (aiding in the computation of static corrections), to mapping depth of crustal basins.
http://pal.boisestate.edu/mediawiki/index.php/Seismic_interferometry
S31A-1899
Equipartition Assessment Using one Station
The elastodynamic Green function can be retrieved from averaging cross correlations of recorded motions within a equipartitioned diffuse seismic field. If only one station is available the autocorrelation allows computing the kinetic energy density and its average is proportional to the imaginary part of the Green function trace at the source. The sum of average energy densities for each degree of freedom (1, 2 and 3) equals the sum of energy densities associated to the various wave types (e.g. P, SV, and SH). For a full elastic space , energy densities fulfill: E1+E2+E3 = EP+ES and equipartition implies that energy is shared in fixed proportions. By means of carefully designed experiments (Hennino et al., 2001), the elastic energy components are separated in both dilatational and shear waves to compute spatial derivatives from small array recordings. It was found that coda waves display elastic equipartition as predicted by theory. Alternatively, here we explore the use of the degrees of freedom of a 3D system to assess the equipartitioned nature of the seismic field using only one station. In the full space case the energies related to each orthogonal direction are equal, therefore each direction shares one third of total available energy density. We used the same data set gathered by Hennino et al. at Chilpancingo, Mexico. Twelve earthquakes were recorded at the four stations of a small array. For each component we calculate the Hilbert envelope. An estimate of the component energies is obtained as fraction of the total energy. For a smoothed running average of fifteen seconds the fraction has a mean close to approximately 0.33, at several portions of records. These preliminary results suggest that equipartition in the Maxwell sense is displayed not only by the coda but by the pre-event noise as well. Another set of data was analyzed. A series of records were obtained at four stations in deep sites in the Tautona and Mponeng gold mines, South Africa. The depth of the recordings ranges from 2 to 3.5 km and the records show some small earthquakes (M=2) and noise as well. A preliminary analysis suggests that the coda exhibits clear elastic equipartition in the frequency band 10 – 30 Hz. Although in some cases the pre- event noise shows equipartition. This feature is not ubiquitous. Our results from analysis of recorded data in Chilpancingo, Mexico, and Tautona Mine, South Africa suggest that equipartition, and thus the diffuse nature of a seismic field, can be estimated using only one station. Acknowledgements. Partial supports from DGAPA-UNAM, Project IN114706, Mexico; and from project DyETI of INSU-CNRS, France are greatly appreciated.
S31A-1900
Data Requirements for the Retrieval of Regional-Scale Reflection Responses by Cross Correlation
With the deployment of large arrays of receivers the crust and upper mantle under an array can be imaged without the occurrence of local earthquakes. Instead, transmitted waves can be used which are caused by earthquakes at teleseismic distances. Either stacks of receiver functions (mode conversions in the transmission responses) or, more generally, the backpropagation of forward scattered energy can be used to image medium contrasts in the crust and upper mantel. The last few years, an alternative technique is under development, Interferometric Seismic Imaging (ISI). With this technique, reflection data are extracted from the coda of transmissions. That is, the reflection response between receivers is retrieved by a summation of cross correlations of responses from several teleseismic earthquakes. Subsequently, the retrieved reflection responses are migrated to obtain an image. With ISI, higher quality images might be obtained than with techniques that use forward scattered energy, since a higher resolution can be reached when imaging reflected amplitudes instead of forward scattered amplitudes. First though, one needs to establish that with ISI, under realistic conditions, a true reflectivity image can be obtained that is not obscured by artifacts. The theory of ISI is well established, but the requirements for ISI are not always met in practice. We discuss what conditions need to be fulfilled to extract good quality reflection data from transmission responses. These conditions are for example related to the number of sources and their position with respect to the array. Even if not all requirements are met, we may still obtain useful information, but the retrieved reflection responses need to be interpreted with knowledge of the omissions. We illustrate the requirements both with synthetic data examples and actual data examples from the Laramie broadband array (2000-2001) and show what artifacts may occur when not all requirements are met.
S31A-1901
Temporal Changes of Seismic Velocity of Shallow Structure Associated With the 2000 Miyakejima Volcano Activity as Inferred From Ambient Seismic Noise Correlation Analyses
Miyakejima Island, which is located about 170 km to the south of Tokyo, Japan, is an active volcano of basaltic magma. In 2000 volcanic activity started with magma ascent and migration northwestwardly on June 26 - 27. Then, the volcano formed a caldera on the summit in July, and large amount of volcanic gas emission continued from late August until now. We analyze the ambient seismic noise recorded at three NIED seismic stations (MKK, MKT, and MKS) in the island in order to study the volcano structure behavior associated with such significant volcanic activities. We apply cross correlation analyses to the continuous records of vertical component of short period seismometers (1 s). The data are sampled at a frequency of 100 Hz with an A/D resolution of 16-bit. We calculate cross correlation functions (CCFs) for time window of 60 s for each station pair. We stack the CCFs for each month and bandpass filter the stacked data at frequency band 0.4 - 0.8 Hz. The stacked CCFs, which may represent the Green function between two stations, at station pairs MKK - MKS (the distance is 1.8 km) and MKT - MKS (the distance is 3.9 km) show wave packets with large amplitudes at both sides (positive and negative time delays). The wave packets propagate at group velocities of about 0.8 - 1.0 km/s. The stacked CCFs for MKK - MKT (the distance is 3.1 km) is one sided (negative time delay). Such asymmetric might be due to the inhomogeneous distribution of propagation direction of ambient seismic noise, so we do not use the data for the following analyses. Comparing the CCFs obtained for periods from July 1999 to June 2000 with that of October 2002, we observe small phase difference of the main wave packet. Our results show that for station pair MKK - MKS, whose path crosses the northern part of the island, velocity increased about 1.6 % after the 2000 volcanic activity. For MKT - MKS, whose path closely crosses the newly formed caldera, we estimate the velocity decrease of about 1.5 %. Such velocity increase and decrease observed at Miyakejima Island might be caused by the stress increase or decrease in the shallow structure due to volcanic pressure sources, volcanic gas permeation in the volcanic edifice, and other phenomena associated with the 2000 volcanic activity.