Swell Noise Suppression by Wiener Prediction Filter
Derman Dondurur
a,
⁎
, Hakan Karslı
b
a
Dokuz Eylül University, Institute of Marine Sciences and Technology, Baku Street, No: 100, Inciraltı, 35340, İzmir, Turkey
b
Karadeniz Technical University, Department of Geophysics, 61080, Trabzon, Turkey
abstractarticle info
Article history:
Received 5 September 2011
Accepted 2 February 2012
Available online 10 February 2012
Keywords:
Low frequency components
Swell Noise
Data processing
Marine Seismics
Wiener prediction filter
The conventional method to remove swell noise from raw seismic data is band–pass filtering (BPF). It ideally
removes whole low frequency band of the spectral content, which results in a total loss of the amplitudes
concerning the low frequency reflections from deeper reflectors, and hence lower resolution in the deeper
reflection events. The procedure described here attenuates swell noise from seismic data while preserving
the reflection amplitudes at low frequency band. The proposed Wiener prediction filter (WPF) method is
used to estimate the swell noise embedded in the raw marine seismic data and then the estimated noise is
subtracted from shots by a trace–by–trace basis. It is observed that the deeper reflections have significantly
higher amplitudes and show better trace–by–trace consistency in the final migration sections obtained by
the WPF application.
The WPF method removes most of the swell noise and may be an alternative filtering technique to the
conventional BPF method. It can be used with high resolution marine seismic data which may have weaker
reflection amplitudes from deeper reflectors. It also improves the lateral continuity of the events which
may be useful for auto–picking tools such as automatic event tracking. We propose that the method can
effectively be used to remove any type of coherent noise providing that a suitable noise model can be deter-
mined from the data itself.
© 2012 Elsevier B.V. All rights reserved.
1. Introduction
Raw marine seismic data are contaminated by various types of
noises, such as operational noise, air gun's bubble effect, multiples, lin-
ear cable and tail buoy noises, guided waves, streamer depth–leveller
noise and swell noise. The primary reflections have generally much
weaker amplitudes than these different coherent or incoherent noise
components. If the noise is separated from the signal by one of its spe-
cific properties, such as frequency band, velocity or amplitude, and
then it is possible to suppress the noise while preserving the signal.
The swell noise is the most dominant noise type on the raw marine
shot gathers and mainly originates from the rough sea state due to the
weather conditions which produce wind–driven longitudinal sea sur-
face waves, and can be regarded as incoherent noise in marine data
sets. Elboth et al. (2009) suggested two key mechanisms for swell
noise generation: (1) hydrostatic pressure fluctuations due to the
vertical motion of the ocean because of the weather conditions, and
(2) dynamic pressure variations along the streamer due to a turbu-
lent layer surrounding the streamer.
The main characteristics of the swell noise are its large amplitude
and low frequency content, and it sometimes induces delays or tem-
porary suspension of data acquisition in marine surveys (Elboth et al.,
2009; Smith, 1999). Sometimes it is so strong that it can obscure un-
derlying primary reflections and degrade overall data quality. The
swell noise and low frequency reflections from deeper reflectors in-
terfere at the low frequency band of the amplitude spectrum. There-
fore, attenuation of the swell noise, while preserving the reflection
amplitudes, is a key problem in the acquisition and early stages of
the processing. A trade–off therefore exists between swell noise at-
tenuation and low frequency signal preservation during the data pro-
cessing. To manage this problem, new techniques based on noise
estimation and separation strategies in attenuating of the noise
have recently been developed and used by many researchers
(Brown and Clapp, 2000; Chiu and Howell, 2008; Guo, 2003; Karslı
and Bayrak, 2004; Nemeth et al., 2000; Özbek, 2000; Schonewille et
al., 2008; Ulrych et al., 1999; Watts et al., 1999).
The most of the methods developed to attenuate high amplitude
swell noise are based on the rejection of low frequency band of the
recorded data, e.g. suppressing the amplitudes at 0–15 Hz frequency
band (Elboth et al., 2009; Yılmaz, 2001). Conventional techniques
based on the frequency domain discrimination of the swell noise
and primary reflections include 1D high–pass (or low–cut) filtering,
band–pass filtering, f–k and f–x filtering (Cambois and Frelet, 1995;
Schonewille et al., 2008; Yılmaz, 2001). Apart from these conventional
methods, Bekara et al. (2008) and Bekara and van der Baan (2010)
sug-
gested an automatic technique for the detection of large amplitude
noise, such as swell noise, and developed an algorithm to suppress it
in f–x domain. In addition, Elboth et al. (2010) applied time frequency
Journal of Applied Geophysics 80 (2012) 91–100
⁎ Corresponding author. Tel.: +90 232 2785565; fax: +90 232 2785082.
E-mail addresses: derman.dondurur@deu.edu.tr (D. Dondurur), hkarsli@ktu.edu.tr
(H. Karslı).
0926-9851/$ – see front matter © 2012 Elsevier B.V. All rights reserved.
doi:10.1016/j.jappgeo.2012.02.001
Contents lists available at SciVerse ScienceDirect
Journal of Applied Geophysics
journal homepage: www.elsevier.com/locate/jappgeo