Instrumentation ViewPoint. Autumn 2005 - …digital.csic.es/bitstream/10261/134554/1/Signal...

Instrumentation ViewPoint. Autumn 2005

62

the future. We are now working on phase-weightedstack algorithms [3], in order to adapt them to theprocessing sequences of wide-angle seismics.

5. AcknowledgementThis work was supported by the projectSigSensual ref. REN2003-08341-C03-C01-02and CTM2004-04510-C03-02, the MarineTechnology Unit, the Public University of Navarraand the Catalan Polytechnical University, SPAIN.I. Rodriguez and C. Simon are financed by I3Pcontracts.

6. References[1] Seismic Unix, Center for Wave Phenomena,C o l o r a d o S c h o o l o f M i n e s ,ht tp : / /www.cwp.mines.edu/cwpcodes/[2] S. Ventosa, C. Simon, M. Schimmel, Signalprocessing techniques applied to seismic signaldetection, Martech, 2005.[3] M. Schimmel, H. Paulssen, Noise reductionand detection of weak, coherent signals throughphase-weighted stacks, Geophysical JournalInternational, 1997.

Signal processing techniques applied to seismic signal detectionS. Ventosa (1), C. Simon (1), M. Schimmel (2)

(1) Unitat de Tecnologia Marina,Pg. Marítim de la Barceloneta 37-49

08003 - Barcelona - SpainEmail: sventosa@utm.csic.es.

(2) Instituto de Ciencias de la Tierra ‘Jaume Almera’08028 - Barcelona - Spain

1. IntroductionWide-angle reflection/refraction seismics (WAS)is one of the most commonly used techniques todetermine the structure of the Earth’s crust. Theacquisition system includes a number of energysources and receivers located at increasingdistances from the sources that record both thereflections and the continuous refractions of theseismic waves propagating into the medium. Dataare represented in the so-called record sections,which display recordings of the different sourcesregistered by a given receiver (or vice-versa) intooffset-travel time diagrams.

Useful seismic phases present in the recordsections are usually masked by different types ofnoise, which may be caused by the instrument(electronic noise, quantification …), by theenvironment (ship noise, cetacean, sea wavecourse, currents …), or even by the signal itself(higher order reverberations and scattered signalswhich obscure later arriving prominent phases).In order to obtain as much of the valuableinformation contained in the record sections aspossible, it is necessary to process the collecteddata in order to either (1) improve the signal tonoise ratio, or (2) remove or attenuate some wellcharacterized phases, such as the water waveor the first reverberations.

Most of existing techniques to improve the signalto noise ratio assume that noise is stationary andGaussian. However, this is usually not the case,so it is necessary to apply alternative techniquesfor an improved signal detection. These methodsare usually based on nonlinear signal processingtechniques that make use of other signal propertiesthat are more robust against outliers, such asmedian estimators.

2. Polarization filtersOne of the techniques used in order to distinguish

signal form noise are filters based on the degreeof polarization [1-2]. [1] and [2] define a signalby an arbitrary but constant polarizationthroughout the course of the signal. This meansthat these filters use the complete wavefield,i.e., the three components of the geophone, tomeasure the variation of polarization alongtraces regardless its category (lineal, circularor elliptic) or direction.

According to the characteristics of the signal toprocess, we have considered two relatedtechniques: one works in the time domain,whereas the other one works in the time-frequency domain through a windowed FourierTransform (e.g., Short Time Fourier Transform,S-Transform). Alternative approaches, suchas the Wavelet Transform can be employed.

Figure 1: Three-component test data and itsdegree of polarization. The polarized signal is

indicated by the grey background.

Instrumentation ViewPoint.Autumn 2005

63

The approach we present here can be separatedin three main steps. First, we estimate theinstantaneous polarization attributes of therecorded signal. In the time-domain method,analytic signal properties are used [3], while inthe time-frequency domain method we usespectral signal properties [4]. The instantaneousparameters are used to compute the degree ofpolarization in a second step, in order to measurethe variation in direction of the polarizationvector, figure 1. Finally, a mask is formed bythis measure and applied to the record section.It allows attenuating noisy parts that are poorlypolarized without affecting polarized areas.

If the density of traces is high then it is possibleto use correlations between different traces andthe directivity of waves to reduce the noise levelby means of spatial averaging filtering [1], figure2.

3. Results and DiscussionAn example of application on real data take fromLigurian Sardinian wide-angle seismic profileacquired during 1995 in the WesternMediterranean, figure 3.

We reprogrammed the filters consideringspecially the calculation speed and memoryusage to enable an efficient processing of hugedata volumes and to include them into theWASPAR (Wide-Angle reflection/refractionSeismic data Processing And Representation)platform [5]. These processing tools developedto date can be used to emphasize some specifictype of waves, remove unwanted phases, suchas water reverberations and other multiples, orto extract a well characterized wave from therest of the record section.

4. AcknowledgementThis work was supported by the projectSigSensual ref. REN2003-08341-C03-C01-02and CTM2004-04510-C03-02. S. Ventosa andC. Simon are financed by I3P contracts.

5. References[1]M. Schimmel, and J. Gallart, “The use ofinstantaneous polarization attributes for seismicsignal detection and image enhancement”,Geophysical Journal International, Vol. 155, pp.653-668. July 2003.

[2]M. Schimmel, and J. Gallart, “Degree ofpolarization filter for frequency-dependent signalenhancement through noise suppression”,Bulletin of Seismological Society of America,Vol. 94, No. 3, pp. 1016-1035, June 2004.

[3]I. B. Morozov, and S. B. Smithson,“Instantaneous polarization attributes anddirectional filtering”, Geophysical JournalInternational, Vol. 61, No. 2, pp. 872-881, May-June 1996.

[4]J. C. Samson, and J. V. Olson, “Somecomments on the descriptions of the polarizationstates of waves”, Geophysical Journal of theRoyal Astronomical Society, Vol. 61, pp. 115-129, 1980.

[5]I. Rodriguez, “A new software tool for Wide-Angle reflection/refraction Seismic dataProcessing and Representation (WASPAR),International Workshop on Marine Technology,Nov 2005.

Figure 2: Spatial averaging window. The windowis a function of time, trace offset and slowness.

Figure 3: Bandpass filtered input data (up), outputof polarization filter (down).