S52A-01
Generalized average of signals (GAS) - a new method for denoising and phase detection
A novel method called Generalized Average of Signals (GAS) was developed and tested during the last two years (Málek et al., in press). This method is designed for processing of seismograms from dense seismic arrays and is convenient mainly for denoising and weak phase detection. The main idea of the GAS method is based on non-linear stacking of seismograms in frequency domain, which considerably improves signal-to-noise ratio of coherent seismograms. Several synthetic tests of the GAS method are presented and the results are compared with the PWS method of Schimell and Paulssen (1997). Moreover, examples of application on real data are presented. These examples were chosen to show a broad applicability of the method in experiments of different scales. The first one shows identification of S-waves on seismograms from shallow seismic. The second one concerns identification of converted waves from local earthquakes registered at the WEBNET local network in western Bohemia. Finally, the third one depicts identification of PKIKP onsets on seismograms of teleseismic earthquakes. Schimmel, M., Paulssen H. (1997): Noise reduction and detection of weak, coherent signals through phase- weighted stacks. Geophys. J. Int. 130, 497-505. Málek J., Kolínský P., Štrunc J. and Valenta J. (2007): Generalized average of signals (GAS) - a new method for detection of very weak waves in seismograms. Acta Geodyn. et Geomater., in press. http://www.irsm.cas.cz
S52A-02 INVITED
Broad-band Measurement and Inversion of Interstation Surface-Wave Dispersion
We present a novel combination of waveform analysis techniques for accurate, broad-band interstation measurement of surface-wave phase velocities. We first measure the fundamental-mode dispersion between a pair of stations by cross-correlating seismograms from the stations (Meier et al. GJI 2004). Waveforms observed at short periods (10-20 s) are complex, distorted by diffraction. It turns out, however, that patterns of waveform complexity often change little as seismic waves propagate from one station to another nearby; cross-correlation can extract accurate measurements even from signals that appear prohibitively complex. We then measure average phase velocities between sources and stations by means of the Automated Multimode Inversion (AMI) of surface- and S-wave forms (Lebedev et al. GJI 2005), and calculate interstation dispersion from each pair of measurements (same event, both stations). AMI synthesizes complete seismograms, and fundamental-mode dispersion can be measured even when the mode interferes with energetic S waves. The cross-correlation technique thus provides short-period measurements and the multimode-waveform technique yields more longest-period measurements, especially for Love waves. Robust dispersion curves with error estimates are derived by averaging over tens or hundreds of smooth curves measured using different events. Any one-event measurement can be biased due to diffraction. Selection of only smooth portions of the curves, removal of outliers, and averaging over many curves obtained with earthquake signals from different regions and from different directions combine to enhance the accuracy of the measured dispersion. Applications of the method to permanent-station data produced measurements in period ranges 10-300 s and broader, with resolving power for isotropic and anisotropic structure from the upper crust to deep upper mantle. S-velocity profiles can be computed from the data by solving small, well-determined inverse problems. Using series of target tests, we can also explore the model spaces of the inverse problems and derive robust ranges of seismic structure parameters consistent with the data. We present results of measurements and inversions for stable and active continental regions, with inferences on structure and dynamics of the lithosphere and asthenosphere.
S52A-03 INVITED
Identifying Higher-Mode Surface Waves From Noise Cross-Correlations
The cross-correlations of seismic noise over long distances offers a new type of data in geophysical studies. Recognizing that the cross-correlation between stations is proportional to the elastic Green's function, we can free ourselves of the resolution problems inherent in illuminating a study area from the uneven distribution of earthquakes: an even distribution of stations over the study area will allow for even resolution. However, because earthquakes occur at various depths, they can excite certain phases much more energetically than those observed in the cross-correlation data, which are dominated by the fundamental-mode surface wave. These higher-mode surface waves and body wave phases are sensitive to deeper structures within the Earth compared to the fundamental mode. Though these higher-mode surface waves and body waves are subdued, they are present in the noise cross-correlations. We use a time-frequency stacking algorithm, based on the S-transform, coupled with a filter designed from the multi-channel coherence across an array, to enhance these faster arrivals. We observe a coherent phase arriving well before the fundamental-mode surface waves that we interpret to be a higher-mode surface wave. Observations like these will be necessary to take advantage of the increased resolution offered by noise cross-correlations input into tomographic inversions while simultaneously placing constraints on upper mantle structure.
S52A-04 INVITED
Ambient Noise Tomography in Western China and Northern India
Surface wave tomography has proven particularly useful in imaging Earth's crust and uppermost mantle on both regional and global scales. However, short-period surface wave dispersion measurements, which are most useful to constrain the structure of the crust and uppermost upper mantle, are extremely hard to obtain from seismic events due to scattering and attenuation. Ambient seismic noise is rich in short-period surface waves from which the empirical Green's functions of surface waves between pairs of stations can be extracted by cross-correlating long noise sequences. Tomography based on surface wave dispersion obtained from the empirical Green's functions has been shown to produce high-resolution, short-period (6-30 s) surface wave dispersion maps across numerous regions of the Earth in both regional and continental scales. This study applies the ambient noise tomography method to the significant data resources that are now available and are continuing to emerge in India and W. China, which will yield dense path coverage in northern India and western China. These resources include data from the permanent Federation of Digital Seismographic Network (FDSN), temporary US PASSCAL installations in and around China, Chinese installations in Tibet, and temporary broad-band deployments in India. The resulting tomography maps provide new information to improve the understanding of regional tectonic processes.
S52A-05
Ambient Noise Surface Wave Tomography Across Europe
We present updated ambient noise tomography (ANT) results from Rayleigh wave group and phase velocities obtained from empirical Green functions produced by cross-correlating long ambient noise sequences between available station pairs across Europe. We use longer time series and introduce stations from temporary networks in an attempt to improve upon the initial ANT results of Europe obtained by Yang et al. (2007) using one year of continuous data from the Virtual European Broadband Seismic Network (VEBSN). In this study, we use 4 years of continuous data (2000, 2001, 2005, 2006) from permanent stations throughout Europe, as well as data from temporary installations and emerging regional networks. Data from 2005-2006 includes three PASSCAL experiments, the North Anatolian Fault (NAF) network in Turkey, the REtreating TRench Extension and Accretion Tectonics (RETREAT) network in Northern Italy, and the Calabria-Apennine-Tyrrhenian/Subduction-Collision- Accretion Network (CAT/SCAN) in Southern Italy, the Spanish National Network, the Israel Broadband Seismic Network, and the Virtual European Broadband Seismograph Network (VEBSN) across Europe. Cross- correlations between stations from the Spanish National Network and the two PASSCAL experiments in Italy improve ray-path coverage in the western Mediterranean. Additionally, cross-correlation pairs between stations from the NAF network in Turkey, the Israel Seismic Network, and permanent stations in Europe improve ray coverage in the eastern Mediterranean region. The earlier time period (2000-2001) is included for improved resolution of the eastern Mediterranean, and utilizes continuous data from two PASSCAL experiments, the Eastern Turkey Seismic Experiment and the Mantle Investigation of the Deep Suture between Eurasia and Africa (Midsea), and other permanent and regional stations throughout Europe. Two-years of stacked cross-correlations are used to obtain Rayleigh wave group and phase velocity dispersion curves, and variation among sets of six- month stacks are used to estimate uncertainty measurements based on seasonal variations. The resulting tomographic images are interpreted in the context of the geophysical characteristics of the crust and upper mantle throughout Europe, and correlate well with known sedimentary basins and regions of thickened crust.
S52A-06
Coherency of seismic noise, Green functions and site effects
The newly rediscovered methodology of cross correlating seismic noise (or seismic coda) to retrieve the Green function takes advantage of the coherency of the signals across a set of stations. Only coherent signals are expected to emerge after stacking over a long enough time. Cross-correlation has a significant disadvantage for this purpose, in that the Green function recovered is convolved with the source-time function of the noise source. For seismic waves, this can mean that the microseism peak dominates the signal. We show how the use of the transfer function between sensors provides a better resolved Green function (after inverse Fourier transform), because the deconvolution process removes the effect of the noise source-time function. In addition, we compute the coherence of the seismic noise as a function of frequency and distance, providing information about the effective frequency band over which Green function retrieval is possible. The coherence may also be used in resolution analysis for time reversal as a constraint on the de-coherence length (the distance between sensors over which the signals become uncorrelated). We use the information from the transfer function and the coherence to examine wave propagation effects (attenuation and site effects) for closely spaced stations compared to a reference station.