You are on page 1of 6

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/261191644

First Arrival Picking in Seismic Refraction Data by Cross-Corelation


Technique

Conference Paper · January 2011

CITATIONS READS

4 3,031

2 authors:

Mustafa Senkaya Hakan Karslı


Uludag University Karadeniz Technical University
32 PUBLICATIONS 87 CITATIONS 54 PUBLICATIONS 298 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Selection of suitable landfill areas and determining of alternative solid waste disposal systems project View project

Işıklar (Trabzon-Maçka) Potansiyel Heyelan Alanının Bütünleşik Çok Boyutlu Jeofiziksel Anatomisi (Integrated Multi-Dimensional Geophysical Anatomy of Işıklar
(Trabzon-Maçka) Potential Land Slide Area) View project

All content following this page was uploaded by Mustafa Senkaya on 31 March 2014.

The user has requested enhancement of the downloaded file.


A22
First Arrival Picking in Seismic Refraction Data by
Cross-Corelation Technique
M. Senkaya* (Karadeniz Technical University) & H. Karsli (Karadeniz
Technical University)

SUMMARY
The quantified interpretation of the seismic refraction record depends on correct and confident picking
first arrival times. However, if the refraction data are poor, traditional first arrival picking techniques are
not enough. For this reason, novel techniques are necessary to obtain first arrival times from such data. In
this study the cross-correlation technique (CCT), commonly used in reflection data processing, was
applied to determine first arrival times. By means of the technique, the process of the first arrival time
picking is realized semi-automatically and erroneous picking to be arise from noises was decreased and
picking procedure was attributed to mathematical base. Also, effectiveness of semi-automatic picking
was especially evaluated with seismic refraction tomography.

6th Congress of Balkan Geophysical Society - Budapest, Hungary


3-6 October 2011
Introduction

The accurate determination of the first-break times is needed for high quality interpretation,
tomographic solutions and calculating the static corrections, a fundamental stage of seismic data
processing. Clearly, the effectiveness of refraction data interpretation depends on the picking-process
reliability. Generally, first-break quality is related to the near-surface structure, source type, and
signal-to-noise ratio (S/N) conditions. Therefore, the automated picking of first breaks can be a very
difficult task if data are acquired in complex near-surface scenarios or if the S/N is low (Yilmaz,
2001). Traditionally, the determination of the first break times was carried out by a visual inspection
of the amplitudes, operator's eye estimation capacity and experience, scaling, quality of data imaging
and waveform changes. These are very time consuming and this strategy can lead to inconsistent
picks because it relies on the picking-operator subjectivity. Moreover, if seismic data include
background noises existing in data acquisition medium, picking first arrivals are directly not simple
and reliable, thus making the picking process even more difficult. As a result, when the data volume is
large and the data quality is poor, the picking procedure can take up to 20–30% of the total processing
time (Sabbione and Velis, 2010).

During the last few decades, many techniques have been developed for determining first break times
automatically or semi-automatically. First attempts were based on the cross-correlation of adjacent
traces to find the delay time between first breaks (Peraldi and Clement, 1972). Hatherly (1982) used
first inflection points and least square methods. Gelchinsky and Shtivelman (1983) and Ervin et al.
(1983) suggested use of correlation properties of signals and the application of statistical criterion of
the estimation of first arrivals time. Gu et al. (1992) obtained first arrivals by ratio of long term energy
to short term energy. Murat and Rudman (1992) suggested to use of neural networks to pick
seismic refraction first break. Unfortunately, when background noise is high and data quality is poor,
these techniques tend to fail and the problem of obtaining consistent and reliable picks in an
automated manner remains unresolved.

In this paper, we propose to apply cross-correlation technique (CCT), which is commonly used in
vibro-seismic reflection data processing to extract reflectors more accurately, to refraction data for
detecting first arrival times semi-automatically. Therefore, an operator or source wavelet is firstly
determined according to noise contents of the data, secondly it is cross-correlated with the data by
trace to trace. Then, first arrival area on the data is marked by manual and finally maximal points of
CC data, which indicate the first break times, are automatically detected. It has been applied to
several synthetic and field data. Tests indicate that methods provide more reliable and faster first
arrival time picking on shot data including intermediate and high random noise level than manual and
automated picking.

Method and Applications

The basis of the proposed method depends on using cross-correlation process. It is well known that
CCT is a measure of similarity or independency of any time series to other time series as a function τ
time lag and is routinely applied to vibro-seismic reflection data to determine reflection wavelets.
When the first arrivals part on the shot data is marked, the cross-correlation process may help us to
determine first break times more reliable and correct than traditional one. For the realization of the CC
process, two variables, data and source wavelet as an operator, have to be known. When the data and
source wavelet are being correlated, correlation function has maximum value where source and data
have similar wave form. The time value corresponding to this maximum correlation value represents
first arrival time.

To test CCT, a synthetic data shown in Figure 1b is calculated by using a modified sine wave in
Figure 1a, characterizing a minimum phase source wavelet, for two layer model. In the model,

6th Congress of Balkan Geophysical Society - Budapest, Hungary


3-6 October 2011
number of receivers is 24, source offset, receiver distance and thickness are respectively 5m, 2m and
10m, and layer velocities are 200m/s and 800m/s. The synthetic data don’t include any noise.

Figure 1c shows correlated section which is calculated by cross-correlation process between the data
in Figure 1b and wavelet or operator in Figure 1a. Now, we easily may determine maximum
correlation values which mean first arrival times, on each channel. In this stage, we firstly should
frame first break area in which the algorithm automatically searches correlation maxima for
determining first arrival times. First arrival times related to maxima are shown in Figure 1d by blue
circles. When compared between determined first arrival times by CCT with theoretical ones, it is
clearly seen that they are same values. This accurate is also provided by chi-square error test with %
95 confidences.
1 0.35 0.35 0.3
Calculated teorical w ave travel times
(a) (b) (c) (d)
0.8 0.3 0.3 Direct w aves travel times
0.25 Refracted w aves travel time
Normalized Amplitude

0.6 (at) 0.25 0.25 First arrival times w ere obtained by Correlation
s(t)=A*sin(2πft)*e 0.2

Time (s)
Time (s)

0.4 0.2 0.2

Time (s)
0.15
0.2 0.15 0.15
0.1
0 0.1 0.1

0.05
-0.2 0.05 0.05 Chi-Square error =1.8124e-005s

-0.4 0 0 0
0 0.02 0.04 0.06 0.08 0.1 5 9 13 17 21 25 29 33 37 41 45 49 5 9 13 17 21 25 29 33 37 41 45 49 0 5 7 9 111315171921232527293133353739414345474951
Time (s) Distance (m) Distance (m) Distance (m)

Figure 1 Two layer refraction data modeling. (a) A minimum phase wavelet, (b) calculated synthetic
refraction data, (c) the data after cross-correlation, (d) theoretical (red squares) and semi-
automatically determined first arrivals times (blue circles) by CCT. Times according to the maximum
correlation values on framed area in (c) represent first arrival times.
0.35 0.35 0.35
(a) (b) (c)
0.3 0.3 0.3

0.25 0.25 0.25

0.2 0.2 0.2


Time (s)
Time (s)

Time (s)

0.15 0.15 0.15

0.1 0.1 0.1

0.05 0.05 0.05

0 0 0
5 9 13 17 21 25 29 33 37 41 45 49 5 9 13 17 21 25 29 33 37 41 45 49 5 9 13 17 21 25 29 33 37 41 45 49
Distance (m) Distance (m) Distance (m)
0.3 0.3 0.3
Calculated teorical wav e trav el times
(d) Calculated teorical wav e trav el times
(e)
Calculated teorical wav e trav el times
(f)
Direct wav es trav el times Direct wav es trav el times Direct wav es trav el times
0.25 Ref racted wav es trav el times Ref racted wav es trav el times Ref racted wav es trav el times
0.25 0.25
First arriv al times were obtained by Correlation First arriv al times were obtained by Correlation First arriv al times were obtained by Correlation

0.2 0.2 0.2


Time (s)

Time (s)
Time (s)

0.15 0.15 0.15

0.1 0.1 0.1

Chi-square error =2.2996e-005 s Chi-square error =0.00020758 s Chi-square error =0.00013074 s


0.05 0.05 0.05

0 0 0
0 5 7 9 111315171921232527293133353739414345474951 0 5 7 9 111315171921232527293133353739414345474951 0 5 7 9 111315171921232527293133353739414345474951
Distance (m) Distance (m) Distance (m)

Figure 2 Application of CCT to noisy data.

We applied the CCT to noisy data as shown in Figure 2. In these modeling, amount of noise was
added to the data in Figure 1b. Three different noisy models as shown in Figure 2 are considered
testing the CCT. The models respectively include random noise (%20, Figure 2a), background noise
(%0.1, Figure 2b) and both random and background noises (%20 and %0.1, Figure 2c). The wavelet
in Figure 1a is used as an operator. First arrival times determined from CCT and theoretical times are
compared in Figure 2d-e-f. Therefore, the results show that the CCT can be applied to field data.

6th Congress of Balkan Geophysical Society - Budapest, Hungary


3-6 October 2011
Figure 3 indicate application results of the CCT to field data. It is clearly seen that picking first arrival
times on the field data is rather difficult by manual or automated. To apply the CCT we need an
operator. It is well known, in conventional seismic refraction field studies, the data are generally
gathered by an uncontrolled source such as a simple sledgehammer or weight drop which produces
minimum phase wavelet. But, frequency content and waveforms of the wavelet may change from shot
to shot. Thus, we have to estimate source wavelet for every shot to use in the CCT. In this study,
estimation of source wavelet or operator is estimated with three different ways. First way is directly
taken as a wavelet from one of high S/N ratio channel of seismic refraction data (Figure 3a), second
way is to use inverse Fourier transform of the smooth average power spectrum of data (Figure 3b),
and third way is acceptable mathematical operator similar to minimum phase source wavelet (Figure
3c and 3d). In this stage, by using these ways, the most appropriate operator to be used in CCT can be
determined.

Generally, seismic refraction data are divided into four categories according to easiness of picking
first arrival times; the data may (1) be clear so that first arrival times are easily picked on all channels,
(2) have contaminated with random noise, (3) have waveform which distorted from trace to trace, and
(4) have onsets of first arrivals not to be picked because of high level background noise, insufficient
source power and complex geology. Our experiences show that an operator for (1) can easily be
determined on a trace with high S/N proportion, smooth spectrum is suitable for (2), and a
mathematical operator according to center frequency content of the data to be picked can be used for
(3) and (4). The application results for the four cases are shown in Figure 3. The green filled circles
indicate first arrival times which are determined by CCT.

Figure 4 shows a comparison between traditional (left) and the CCT (right) for tomographic
interpretation of a refraction data which includes six shot and were gathered in noisy environment
where is near to airport and railway. The only difference in obtaining of tomography results in Figure
4a and Figure 4b is first arrival times used, other parameters are same. The result in Figure 4b is more
consistent with geological information for data gathering area. Therefore, the results explain the
importance of accurate determining of the first arrival times.
0.5 0.25 0.25
0.25
(a) (b) (c) (d)
0.4 0.2 0.2
0.2

0.3 0.15 0.15 0.15


Time (s)

Time (s)
Time (s)

Time (s)

0.2 0.1 0.1 0.1

0.1 0.05 0.05 0.05

0 0 0 0
2 4 6 81012141618202224 2 4 6 81012141618202224 7 911131517192123252729 7 911131517192123252729
Distance (m) Distance (m) Distance (m) Distance (m)
Figure 3 Field data examples. The data are (a) clean, (b) with random noise, (c) with distorted
waveform, and (d) with noisy and distorted waveform. Filled green circles indicate first arrival picks
determined by the CCT.

Figure 4 A comparison between traditional (left) and the CCT (right) for tomographic interpretation
of a refraction data.

6th Congress of Balkan Geophysical Society - Budapest, Hungary


3-6 October 2011
Conclusion

In this paper, a new procedure, CCT, has been developed to pick first arrival times for reducing the
manual effort and automatic difficulty in the picking process and accurately perform the picking. The
CCT has been tested on the noiseless or noisy synthetic and field data. The CCT is robust for noisy
data and provide accurate and consistent picks even under the presence of high amplitude noise, bad,
pulse changes, and indistinct first breaks. Besides, the CTT is computationally efficient and easy to
apply because the user needs to estimate only a operator, depending on the noise content of the data.
Results show that the performance of the CCT considerably exceeds the performance of the
traditional one.

Acknowledgements

This work was funded by Karadeniz Technical University-Scientific Research Foundation (Project
code no: 2009.112.007.2).

References

Ervin, C. P., McGinnis, L. D., Otis, R. M. and Hall, M. L. [1983] Automated analysis of marine
refraction data: a computer algorithm. Geophysics, 48, 582-589.

Gelchinsky, B. T. and Shtivelman, V. [1983] Automatic picking of first arrivals and parameterization
of traveltime curve. Geophysical Prospecting, 31, 915-928.

Gu, H., Zhou, H. and Zhang, X. [1992] Automatic pick of first arrival time. Geophysical &
Geochemical Exploration, 16, 120-129.

Hatherly, P. J. [1982] A Computer method for determining seismic first arrival times.
Geophysics, 47, 1431-1436.

Murat, M., and Rudman, A. [1992] Automated first arrival picking: a neural network approach.
Geophysical Prospecting, 40, 587–604.

Peraldi, R., and Clement, A. [1972] Digital processing of refraction data: Study of first arrivals.
Geophysical Prospecting, 20, 529–548.

Sabbione, J. I. and Velis, D. [2010] Automatic first-breaks picking: New strategies and algorithms.
Geophysics, 75, V67-V76.

Yilmaz, O., [2001] Seismic data analysis: Processing, inversion, and interpretation of seismic data.
Society of Exploration Geophysicists, Tulsa.

6th Congress of Balkan Geophysical Society - Budapest, Hungary


3-6 October 2011

View publication stats

You might also like