You are on page 1of 9

291

Int. J Sci. Emerging Tech Vol‐5 No 5 May, 2013

Using Continuous Wavelet Transform and Short


Time Fourier Transform as Spectral
Decomposition Methods to Detect of
Stratigraphic Channel in One of the Iranian
South-West Oil Fields
E. Shokrollahi#1, G. Zargar#2, M.A. Riahi*3
# Petroleum exploration engineering department, Petroleum University of Technology, Abadan, Iran
1 erf_shokrollahi@yahoo.com
2 zargar@put.ac.ir
*
(Institue of Geophysics,University of Tehran, Tehran, Iran)
3 mariahi@ut.ac.ir

Abstract--Classical seismic methods for characterization


order to obtain more geological information. Based on
of hydrocarbon reservoirs have been used for decades. definition, SD refers to any method that produces
Stratigraphic events affect seismic sections in addition to continuous analysis of amplitude, frequency, phase,
structural events. Today the necessity of usage of seismic and energy spectrum. In this method, the output of
sections to determine reservoir extension and each seismic trace is an amplitude spectrum of
stratigraphic characteristics is increasing. Conventional frequency content of input trace that refers to temporal
seismic sections cannot display most of the important center of sample. Therefore, the result of SD of a
events in reservoir studies. The study of frequency seismic trace is a time-frequency analysis window.
content of seismic sections can provide better
understanding to geoscientists, as well as interpreters in One of the oldest methods in obtaining local
the field of petroleum engineering studies. An is spectrum is Discrete Fourier Transform (DFT) which
frequency process is a powerful tool in terms of reservoir is a classical frequency decomposition method. This
imaging. One of the methods that provides is frequency transform calculates the relative intensity of each
slice is Spectral Decomposition (SD). frequency component of the whole signal, but does not
SD is a comprehensive method that eliminates most provide any information about frequency content
of limitations encountered in seismic data to reveal changes with time. Therefore, DFT method is not
geological information. SD provides continuous analysis suitable for non-stationary signals and cannot
of amplitude, frequency, phase, and energy spectrum. determine frequency changes with time (seismic
Therefore, SD is applied to obtain an amplitude waves are not stationary, i.e. Their frequency content
spectrum of frequency content of seismic trace which is
varies with time).
attributed to the temporal center of the sample trace, so
higher frequency resolution is obtained at lower To resolve this problem of DFT, Short Time
frequencies and higher time resolution at higher Fourier Transform (STFT) or Fast Fourier Transform
frequencies. (FFT) has been widely used to decompose non-
In this paper, Short Time Fourier Transform stationary signals. The basic idea in STFT is to divide
(STFT) and Continuous Wavelet Transform (CWT) as a non-stationary signal to small segments (which are
SD methods were used in order to detect stratigraphic considered as stationary parts) and calculate Fourier
channel at one of the Iranian South-West oil fields for transform for each segment. To perform this division,
which it has the potential to be considered as a reservoir. a window function is chosen and multiplied by the
signal. One of the setbacks of this method is the
Keywords: Continuous Wavelet Transform (CWT), Short constant-length time interval obtained from window
Time Fourier Transform (STFT), Spectral Decomposition function. Short temporal window provides suitable
(SD), stratigraphic channel, Sarvak. time resolution but the resulting frequency resolution
is low. If a larger temporal window is selected,
1. Introduction frequency resolution will be improved but time
resolution will be reduced. Therefore, if the goal is
Different methods have been introduced for SD of exact identification of separate time events, methods
seismic data. The goal of all these methods is to in which the window length is automatically adjusted
decompose the seismic signal to its components in to frequency should be used.
__________________________________________________________________________ Decomposition by Continuous Wavelet
International Journal of Science & Emerging Technologies Transform (CWT) is a suitable method to overcome
IJSET, E‐ISSN: 2048 ‐ 8688
Copyright © ExcelingTech, Pub, UK (http://excelingtech.co.uk/)
the resolution problems in decomposition by STFT
method. In CWT the signal is multiplied by a function
292
Int. J Sci. Emerging Tech Vol‐5 No 5 May, 2013

such as window function in STFT in which the length frequency covering length tends towards high
of the window is not constant. The length of temporal frequencies and low frequencies respectively;
window changes according to the required frequency therefore, by increasing frequency resolution, time
resolution. This change in window length is the result resolution decreases consequently and vice versa [5].
of a window function called “wavelet function”. An As hydrocarbon reservoirs are detectable in low
important step in CWT method is selection of the frequencies and thin layers with high temporal
wavelet function which is called Mother wavelet. resolution are detectable in high frequencies this
Commonly used wavelets in CWT are Morlet wavelet, method is a proper method in reservoir
Gaussian wavelet and Mexican-Hat wavelet. Although characterization [6].
there is no absolute optimal wavelet among the
Wavelet transform was first used by Morlet et al.
common wavelets, by considering the purpose and
in 1980s for assessing of seismic data [7].
type of usage, specifications for suitable Mother
wavelet must be considered to choose an optimal Wavelet with symbol (t) is usually defined as a
wavelet. function (t) ϵ L2(R) with zero mean which can be
determined in two time and frequency domains. By
As an example side lobes of Morlet wavelet
reduce vertical resolution of CWT [1]. translating this wavelet (t) a family of wavelets can
be produced:
In this study, performance of CWT and STFT in
mapping stratigraphic channel in one of the Iranian
South-West oil fields using 3-D seismic data, is (2)
, √
investigated.
In which τϵR, σ ≠ 0, σ is scale parameter, τ
2. Theory and Method Translation parameter and is complex conjugate
of (t). As wavelets are usually normalized, a
2.1 Spectral Decomposition Methods coefficient 1/√σ is used. CWT is defined as the inner
product of the signal u (t) and a group of wavelets σ,
2.1.1 Short Time Fourier Transform (STFT) τ (τ). Its formula is as:

Peyton et al. (1998) [2], partyka et al. (1999) [3],


and Marfurt and Kirlin (2001) [4] are the prominent , , ,
founders of this method. ∞

(3)

The basis of this method is selecting an optimal
window length and making its convolution with
Gaussian window, so that the Fourier Transform of In , which is the time-scale map (i.e. the
this window can be calculated. Proper selection of the scalogram).
window is the main problem of this method.
The scalogram cannot provide a direct and clear
Based on the selected temporal window the STFT interpretation of frequency. For interpretation of the
provides a time-frequency spectrum. In STFT, time- time-scale map in terms of a time-frequency map, a
frequency resolution is fixed over the entire time- constant (c), which is obtained from the relation
frequency space by preselecting a window length. between scale and central frequency, must be
Therefore, resolution in seismic data analysis becomes calculated approximately as follows:
dependent on user-specified window length.
(4)
Mathematically, STFT is the inner product of
signal and a time shifted window in time τ
and the frequency ω as shown in ‘Eq (1)’: The easiest step in conversion of scale to
frequency is equating it with a scale-frequency map of
wavelet.
,
∞ The useful wavelets commonly used in wavelet

̅ (1) transform are Morlet, Gaussian and Mexican-Hat. In
this article, these three wavelets are used in CWT
analysis.
Where ̅ is the complex conjugate of .
This method is a useful operator in studying and 2.2 Using CWT and STFT for Practical Study
detection of thin layers. of the Field
2.1.2 Continuous Wavelet Transform (CWT) In this paper, CWT and STFT are used as SD
CWT in signal analysis is considered an methods for detection of the channel in seismic data in
alternative method for frequency distribution analysis one of the Iranian oil fields.
of non-stationary time series. In CWT method by
increasing or decreasing temporal covering length,
293
Int. J Sci. Emerging Tech Vol‐5 No 5 May, 2013

This oil field with a trend of S-N is located in the Figure 3 shows the seismic amplitude response in
Dezful Embayment in the South-West of Iran. Dezful time slice of 1800ms obtained from 3-D seismic cube
Embayment contains major oilfields of Iran, in other of figure 1. This horizon slice is taken 20 ms under the
words, it is the center of South-West Iran oil province, Sarvak horizon shown in figure 2.
which is geologically a part of Northern Arabian plate.
This region, with a flat and even topography, is
covered with a layer of quaternary alluvial sediments
which has a thin thickness.
The geological studies of this oil field show that
in the section related to Sarvak Formation, channel
events commonly exist. Sarvak Formation
stratigraphic ally locates in the second sequence of
Middle cretaceous and has reservoir potential.
Figure 1 shows 3-D seismic cube of the studied
oil field and figure 2 shows the interpreted Sarvak
horizon. The studied section by SD methods for
detection of channel is beneath this horizon.

Figure 3. Seismic amplitude response of time slice


1800 ms (20 ms below the Sarvak horizon)
As it is seen, no specific geological event which
causes changes in seisimc structure is detectable.

CWT has been calculated as follows.CWT was


calculated as isofrequency seismic sections based on
all three Morlet,Gaussian and, Mexican-Hat wavelets
at frequencies of 20, 30, and 40Hz and the results have
been shown in figures 4 to 6 for a temporal depth of
1800ms.

At frequency of 20Hz in figure 4, both Morlet and


Figure 1. 3-D seismic volume of the studied oil field Gaussian wavelets provide better resolutions
compared to Mexican-Hat. At 30Hz frequency in
figure 5, Morlet wavelet provides better resolution
than Gaussian wavelet which in term provides better
amplitude contrast than Mexican-Hat. In 40Hz
frequency in figure 6, Gaussian wavelet provides
relatively a good resolution, although Morlet wavelet
does not provide relatively good resolution, it provides
higher amplitude contrast compared to low resolution
of Mexican-Hat wavelet.
The minimum and maximum seismic amplitudes
as a result of CWT method are shown in table 1.
The obtained results from CWT are generally as
follows:
 In 20 and 30Hz frequencies Morlet,
Figure 2. Interpreted Sarvak horizon from 3-D
Gaussian, and Mexican-Hat wavelets performed better
seismic cube
respectively.
 In 30Hz frequencies Gaussian, Morlet and
Mexican-Hat wavelets performed better respectively.
294
Int. J Sci. Emerging Tech Vol‐5 No 5 May, 2013

(a)

(a)

(b)

(b)

(c)
Figure 4. Illustration of isofrequeny slices of 20Hz
obtained from CWT method. (a) Morlet wavlet, (b)
Gaussian wavelet, (c) Mexican-Hat wavelet (c)

Figure 5. Illustration of isofrequeny slices of 30Hz


obtained from CWT method. (a) Morlet wavlet, (b)
Gaussian wavelet, (c) Mexican-Hat wavelet
295
Int. J Sci. Emerging Tech Vol‐5 No 5 May, 2013

(c)
(a)
Figure 6. Illustration of isofrequeny slices of 40Hz
obtained from CWT method. (a) Morlet wavlet, (b)
Gaussian wavelet, (c) Mexican-Hat wavelet

(b)

Table 1. Estimation of seismic amplitude of isofrequency sections obtained from CWT method

Morlet Wavelet GaussianWavelet Mexican-Hat Wavelet

Max Min Max Min Max Min


amplitude amplitude amplitude amplitude amplitude amplitude
20Hz
isofrequency 0.557274 0.012638 0.166242 0.004634 0.342124 0.010825
sections

30Hz
isofrequency 0.646286 0.020745 0.245229 0.008189 0.37163 0.011042
sections

40Hz
isofrequency 0.603053 0.17234 0.288686 0.007864 0.352619 0.009143
sections
296
Int. J Sci. Emerging Tech Vol‐5 No 5 May, 2013

STFT is the second method of SD used in this


study to obtain isofrequncy seismic sections.In the
following STFT calculation is studied. STFT method
was calculated as isofrequency seismic sections in
frequencies of 15, 25, and 35Hz and in 10, 56, and
100ms time intervals and the results were shown in
figures 7 to 9 for a temporal depth of 1800ms.

Analyzing isofrequency images in figure 7 and 8


shows that isofrequency slices of the temporal
window of 56 ms (figure 7-b and 8-b) illustrate the
thickness and boundaries of the channel much better
than the other two temporal windows of STFT method.
Also 35Hz isofrequency image of 100ms temporal
window shows the channel details better than 56 and
10ms temporal windows of STFT method in figure 9.
(a)
As shown in figures and observed in the
maximum and minimum amplitude values of sections
in table 2, isofrequency sections of STFT have a lower
quality than CWT method and CWT displayed a better
performance compared to STFT method in channel
detection.

The obtained results from STFT are generally as


follows:

Comparison of the obtained results from different


temporal windows in an isofrequency seismic section:
 In 15 and 25Hz isofrequency sections, 56, 100,
and 10ms time intervals performed better in
channel detection respectively. (b)
 In 35Hz isofrequency sections, 100, 56, and
10ms time intervals performed better in channel
detection respectively.
Comparison of isofrequency seismic sections in
equal temporal windows:
 In 10ms temporal window, a better channel
resolution is obtained in frequencies of 35, 25, and
15Hz respectively.
 In 56ms temporal window, a better channel
resolution is obtained in frequencies of 25, 15, and
35Hz respectively.
 In 100ms temporal window, a better channel
resolution is obtained in frequencies of 25, 35, and
15Hz respectively.
(c)

Figure 7. Illustration of isofrequeny slices of 15Hz


obtained from STFT method. (a) 10 ms temporal
window, (b) 56 ms temporal window, (c) 100 ms
temporal window
297
Int. J Sci. Emerging Tech Vol‐5 No 5 May, 2013

(a)
(a)

(b)
(b)

(c)
(c)

Figure 8. Illustration of isofrequeny slices of 25Hz Figure 9. Illustration of isofrequeny slices of 35Hz
obtained from STFT method. (a) 10 ms temporal obtained from STFT method. (a) 10 ms temporal
window, (b) 56 ms temporal window, (c) 100 ms window, (b) 56 ms temporal window, (c) 100 ms
temporal window temporal window
298
Int. J Sci. Emerging Tech Vol‐5 No 5 May, 2013

Table 2. Estimation of seismic amplitude of isofrequency sections obtained from STFT method

10ms temporal window 56ms temporal window 100ms temporal window

Max Min Max Min Max Min amplitude


amplitude amplitude amplitude amplitude amplitude
15Hz
isofrequency 3.06994 0.09092 34.37968 1.485297 31.463892 1.115812
sections

25Hz
isofrequency 4.39585 0.14683 52.11539 1.3913 52.763969 2.252071
sections

35Hz
isofrequency 5.84562 0.16486 58.22383 1.960338 82.458725 3.277157
sections

In figure 10, the position of the studied channel is


shown in vertical seismic section.

Vertical seismic section in figure 10-a is located


where the channel main body is existed corresponding
to Cross line number 5570 of seismic cube. Relative
horizon to Sarvak Formation is characterized in figure.
The oval shows the location of the main body of the
channel. Figures 10-b shows 20Hz is frequency
sections resulting from CWT method using Morlet
wavelet, corresponding to time slice of 1800ms, in
addition to the location of corresponding vertical
section.

(b)

Figure 10. (a) Vertical seismic section


corresponding to Crossline number 5570. Sarvak
Horizon is specified in the figure. The oval shows the
location of main branch of the channel. (b) 20Hz
isofrequency section of CWT method, using Morlet
wavelet corresponding to 1800ms time slice. Red line
shows the location of vertical section shown in figure
10-a
3. Conclusion
Spectral Decomposition is a powerful tool in
detection of channels by using 3-D seismic data.
(a) Application of CWT and STFT as two effective
SD-based methods is proper tools in imaging the
buried channels which provide the possibility of
comparison between the resulted is frequency slices
for these methods.
The best result in channel detection of all CWT is
frequency sections is obtained from Morlet Mother
Wavelet and in 20Hz section. The best result of all
299
Int. J Sci. Emerging Tech Vol‐5 No 5 May, 2013

STFT is frequency sections is obtained in 25Hz is [2] Peyton, L., Bottjer, R., and Partyka, G., 1998,
frequency section of 56ms temporal window. Interpretation of incised valleys using new 3-D
seismic technique: A case history using spectral
Based on the obtained results of this paper, is
decomposition and coherency: The Leading
frequency slices of CWT method relatively showed
Edge, 17, 1294-1298.
better quality than STFT method and detection of
channel with more details is provided well by CWT [3] Partyka, G., Gridley, J., and Lopez, J., 1999,
method. This is because of constant Time-Frequency Interpretational applications of spectral
resolution in STFT method over the entire Time- decomposition in reservoir characterization. The
Frequency space by preselecting a window function. Leading Edge, 18, 353–360.
Contrary to STFT in CWT method, the adaptive
[4] Marfurt, K.J., and Kirlin, R.L., 2001, Narrow
wavelet causes the production of optimum window
band spectral analysis and thin bed tuning: 66,
length relative to frequency content of seismic data. In
1274-1283.
other words in CWT method by increasing or
decreasing temporal covering length, frequency [5] Mallat, S., 1999, A wavelet tour of signal
covering length tends towards high frequencies and processing, 2nd ed.: Academic Press Inc.
low frequencies respectively.
[6] Sinha, S.K., Routh, P.S., Anno, P.D., Phillips, C.,
and Castagna, J.P., 2005, Spectral Decomposition
References of seismic data with continuous-wavelet
transform: Geophysics, 70, 19-25.
[1] Castagna, J.P., and Sun, S., 2006, Comparison of
spectral decomposition methods: First Break, 24, [7] Grossmann, A., and Morlet, J., 1984,
75-79. Decomposition of Hardy functions into square
integrable wavelets of constant shape. SIAM J.
Math. Anal., 15, 723-736,1984.

You might also like