En  |  Pyc

SNDP and Small Aperture Arrays

General provisions.

The experience of exploitation of Small Aperture Seismic Arrays (SASA) have shown that just these instruments are the best ones for detection and location of weak high frequency earthquakes and explosion sources based on seismograms with small signal-to-noise ratio. Small aperture arrays compose the backbone of alpha-stations of International Monitoring System (IMS) for CTBT monitoring.

A conventional SASA may consist of 9-11 stations within the circle of a diameter 2-3 km and a central broadband station; it's better to have 3-4 stations 3-component broadband ones and the rest can be broadband or 1-component high frequency instruments. The aperture of such array provides the high coherence of the regional seismic signals at the sensor outputs in frequency band up to 10 Hz. This allows enhancing signal-to-noise ratio by beamforming procedure and to detect and locate weak seismic events with magnitude 2-2.5 at distances up to 2000-2500 KM.

Deployment of 3-component broadband stations in small aperture array gives the chance to analyse a complete seismogram wavetrain: to recover shear and surface waves from earthquakes and explosions and to identify event source using the features of the wave phases.

Small aperture array can also serve as effective tool for investigation of medium inhomogeneity in a vicinity of the array. Using 3 component sensors allows applying highly efficiently, for example, receiver function analysis, by suppressing the noise at each component. The array processing methods such as semblance and F-K analysis allow to locate strong scatterers around the array. The design of the array as mobile one with radio telemetry links from sensors to central station (local data processing center) would allow to perform several experiments with array installations around different sites of the region. As the result the large area could be studied in respect of detection and location of strong seismic scatterers.

Research and observations conducted last years have shown high performance of small aperture seismic arrays for study of intensity, spectral content and dynamic characteristics of seismic noise and it was found that seismic arrays may be used as the new high-precision tool of measurement of seismic waves velocities and effective Q-factor.

Another frequent application of SASA is seismic tomography, for example, for identification of shallow and deep reflectors in volcanic areas, and so called emission tomography (method based on semblance) for geo-exploration works (search for hydrocarbonate deposits, underground cavities investigation, etc.).
The idea of using precise beamforming procedures for telemetered regional seismic networks to improve location of small events attracts seismologists but it's realization is restricted by next disadvantage: as a rule, these networks are installed in mountain region with strong inhomogeneity of the upper crust beneath the aperture of the network. This leads to poor waveform coherence for seismograms from teleseismic and distant regional events.

The mountain topography provides the high level of seismic noise due to influence of wind and water streams. The noise at different sensors of the network is also incoherent. For these reasons it is impossible to implement the methods of array data processing while analysing network multichannel data. In particular the beamforming, F-K and semblance techniques are not effective for study of such data, especially for high frequency band. This is serious obstacle to use the networks as a tool for monitoring of weak earthquakes and nuclear tests at regional distances. In case of using even conventional beamforming procedures for SASA data, it's possible to significantly improve location of small events. As NORSAR reported, just one SASA gives a location accuracy up to 30 km at a distance of 1500 km. Three Scandinavian SASA improves this number to 15 km.

Certainly using single SASA cannot be a universal case for solving all seismological problems. For instance, focal mechanism may not be determined in this case, same as precise origin depth. A comprehensive solution as we can see it might be found using 3 SASA, each consisting of 3-4 3-component stations. Each SASA will supply with it's information Locsat - very efficient seismic source location algorithm - providing high location precision (assuming availability of regional travel time tables), and all other conventional seismological procedures will also be available, such as focal mechanism determination. A cost of such installation will be higher than a cost of one SASA with the same number of stations, but will provide wider seismological facilities. Also, it's cost will be lower (taking into account also installation and maintenance) than large aperture regional network.

Algorithms and software.

The use of small aperture array recordings allow to strongly enhance the signal-to-noise ratio due to implementation of special adaptive data processing procedures suppressing a seismic noise. We call these procedures adaptive group filtering (adaptive beamforming) algorithms. These algorithms are multichannel filters which vector frequency responses essentially utilize the coherence features of a seismic noise. Latter is especially strong in the case of array location near the ocean cost, because in this case the noise is mainly constituted of seismic surface waves generated by sea surfs.

The spatial and time spectra of the noise waves are dependent on storms areas in the ocean and vary in time. So effective suppression of coherent noise components is possible only if adaptive array data processing procedures are implemented which periodically estimate the matrix power spectral density of array noise recordings. We implement statistically and computationally effective group filtering algorithms and programs, which practically eliminate the coherent noise components and retain seismic phase waveforms undistorted.

SNDP utilizes this advanced technique for regional seismic monitoring in areas affected by strong seismic noise. This concern is connected with modern tendency in seismology to deploy seismic stations at noisy areas such as sea shores, ocean islands and industrial areas for monitoring of weak earthquakes and underground explosions. We mainly emphasize on seismic installations on sea-shores.  Since such observational sites are affected by strong seismic noise generated by sea waves, it makes it difficult to process signals from weak seismic events by the conventional means and to provide acceptable magnitude threshold for regional seismic events reliably detected and located.

To gain all the advantages of SNDP, data from Small Aperture Seismic Array(s) must be used. SNDP provide more reliable detection and accurate parameter estimation of various seismic wave phases from small regional events in conditions of strong non-stationary seismic noise stipulated by sea and mountain environment (water streams, wind attacks and so on). SNDA software provides decreasing of the magnitude threshold for reliably detected and located regional events based on the data from single SASA, in comparison with conventional software implemented currently at the International Data Center (IDC) and National Data Centers (NDC) of the CTBT International Monitoring System (IMS).

The software includes the special adaptation algorithms and programs that provide the estimation of noise statistical characteristics by multidimensional autoregressive - moving averaged modeling of current array noise. The high coherency of seismic noise generated, for example, by sea-shore waves gives one a chance to eliminate it by filtering the array data with a multichannel Wiener filter adapted for current noise statistical characteristics.

This procedure can be regarded as statistically optimal beam-forming. Suppression of the coherent noise component can provide a significant signal-to-noise ratio gain and enhance reliability of seismic wave phases detection and accuracy of their parameter estimation. Evaluation of wave phase onset uses a maximum likelihood procedure that estimates a moment when power spectral density of a beam trace is abruptly changed due to wave phase arrival.

For estimation of azimuth and apparent velocity of wave phases the original procedures were implemented. The algorithms have adaptive structure and give the possibility to estimate properly an arrival direction of a weak seismic phase on a background of interfering wave, for example, coherent coda of previous strong wave phase.

For adaptive signal detection we use a statistically optimal algorithm for testing the hypothesis about change of power spectral density (PSD) of a time series being observed. Estimation of wave phase onset times is made by the maximum likelihood procedure for evaluation of a moment when power spectral density of a beam trace is abruptly changed due to wave phase arrival.

For estimation of azimuth and apparent velocity of wave phases, we implement the original adaptive maximum likelihood procedure, which gives the possibility to estimate arrival directions of weak seismic phases on a background of interfering waves, for example, coherent coda waves of a previous strong phase. The final step of the processing is a location of seismic event epicenters on the basis of a single seismic array data, as well as on the basis of data from many SASA, or from a telemetered network.

The SNDP also includes a subsystem for automated source identification which implements different seismogram identification features, including new discrimination approach involving calculation of seismic source moment tensor as function of frequency. The subsystem uses conventional statistical classification rules and modern neural networking algorithms with selection of the most informative subset of identification features from the total set, and consistent estimation of identification error probability in the region based on learning collections of earthquake and explosion seismograms.

The usefulness of the developed algorithms has been proved by simulation studies and by testing on data from small aperture Scandinavian arrays NORESS, ARCESS and FINESA.

The system was successfully represented and tested in NORSAR in 1995, and later was successfully deployed in several research institutions and governmental agencies, including Phillips Laboratory, USAF.

Your are here: