Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
a r t i c l e i n f o a b s t r a c t
Article history: Vibroseis is a source used commonly for inland seismic exploration. This non-destructive source is often used
Received 16 August 2012 in urban areas with strong environmental noise. The main goal of seismic data processing is to increase the
Accepted 4 February 2013 signal/noise ratio where a determinant step is deconvolution. Vibroseis seismic data do not meet the basic
Available online 24 February 2013
minimum-phase assumption for the application of spiking and predictive deconvolution, therefore various
techniques, such as phase shift, are applied to the data, to be able to successfully perform deconvolution of
Keywords:
Data processing
vibroseis data.
Seismic This work analyzes the application of deconvolution techniques before and after cross-correlation on a real data
Vibroseis deconvolution set acquired for high resolution prospection of deep aquifers. In particular, we compare pre-correlation spiking
uncorrelated data and predictive deconvolution with Wiener filtering and with post-correlation time variant spectral whitening
deconvolution. The main result is that at small offsets, post cross-correlation spectral whitening deconvolution
and pre-correlation spiking deconvolution yield comparable results, while for large offsets the best result is
obtained by applying a pre-cross-correlation predictive deconvolution.
© 2013 Elsevier B.V. All rights reserved.
0926-9851/$ – see front matter © 2013 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.jappgeo.2013.02.009
L. Baradello, F. Accaino / Journal of Applied Geophysics 92 (2013) 50–56 51
Fig. 1. Position map of the seismic line (base map from Google Earth, December 2012). The geological map is superimposed. The station number is also indicated.
wavelet with mixed-phase and only if the earth attenuation is low, the aquifers, while the second is characterized by alternating layers of
data can be approximated to a zero-phase (Pritchett, 1990). sand, gravel and clay, hosting important multi-layered confined aqui-
Over the years, several authors analyzed the deconvolution problem fers (see Fig. 1).
of mixed-phase data and tried several approaches and solutions. Gibson
and Larner (1984) converted the Klauder wavelet to a minimum-phase
equivalent and used minimum-phase deconvolution to remove both
the Klauder wavelet and the earth attenuation, leaving a zero-phase re-
sult. Cambois (2000) has not used the pre-whitening method as Gibson
and Larner but adjusted the low frequencies of the sweep, obtaining a
good minimum-phase equivalent. Porsani and Ursin (1998) created
an algorithm for mixed‐phase deconvolution valid only for pulses
whose Z-transform has no zeros on the unit circle. Carlini and Storer
(1991) tried a deterministic and statistic approach to process the phase
signal. Tygel et al. (1991) solved the problem of extracting a mixed‐
delay source wavelet from a point‐source seismogram without any fur-
ther assumptions about the source pulse or the model parameters.
In this paper, we investigate the potential of the deconvolution on
the uncorrelated data to remove the noise that was already discussed
by Ristow and Jurczyk (1975). They compared deconvolution opera-
tors (calculated with Wiener method) before and after correlation.
They showed that this method reduces mono-frequency disturbs of
unwanted oscillating events. We applied this method to a high reso-
lution seismic line investigating deep aquifers located close to the
water spring line of the Friuli Plain (North-East Italy). The water
spring line is the area where the leakage of water from the water
table of the upper Friuli Plain occurs, due to a rapid decrease of aver-
age permeability. Actually the water spring line divides the Friuli
Plain into two areas: the upper and lower Friuli Plain. The first is char-
acterized by the presence of gravel deposits, which host unconfined
Table 1
Acquisition parameters.
Sample rate 1 ms
Sweep length 10 s
Record length 13 s
Frequency range 10–250 Hz
Shot interval 10 m
Fig. 2. Example of uncorrelated (A) and correlated shots with ground force (B). The
Receiver interval 10 m
black arrow indicates the factory noise, while the white arrow indicates an example
Channels 160
of traffic noise. Trace interval is 10 m.
52 L. Baradello, F. Accaino / Journal of Applied Geophysics 92 (2013) 50–56
Fig. 3. Autocorrelation of a near offset trace (A) and of the ground force (B).
The stratigraphy of the area is composed by Miocene Molasse de- The source was a MiniVib IVI T-2500 mounted on a small truck, weighing
posits overlaid by more than 500 m thick Plio-quaternary sediments 65,000 kg and capable of generating a frequency modulated signal from
(Nicolich et al., 2004). Two shallow aquifers are located around 30 and 10 to 550 Hz. A linear up-sweep from 10 Hz to 250 Hz with a vibration
180 m depth, and a deep aquifer is present at about 500 m (Giustiniani time of 10 s in addition to a taper of 200 ms was used. The taper is applied
et al., 2008), all within the Plio-quaternary sequences. to reduce Gibbs' oscillations. The receiver and shot interval was 10 m,
The seismic line object in this study (Fig. 1) starts in the North-East with a single vibration per VP, for a maximum fold of 80. The data were
where the subsoil is well known (Giustiniani et al., 2009) and the envi-
ronmental noise is small, and ends 1.5 km South-West where the envi-
ronmental noise is high, generated by road traffic and by the activity of
some factories.
We compare the results from a standard processing after cross-
correlation with results obtained by deconvolution applied before
cross-correlation. Both statistical (spike and predictive) and an in-
verse Wiener filter (using the sweep) deconvolutions were also ap-
plied. The final stack sections obtained with different processing are
also shown and discussed.
2. Methods
Fig. 4. Uncorrelated shot after the application of predictive deconvolution. The figure Fig. 5. Comparison of the selected shot after the application of spiking (A), and predictive
must be compared with Fig. 2A. Trace interval is 10 m. deconvolution (B) before cross-correlation. Trace interval is 10 m.
L. Baradello, F. Accaino / Journal of Applied Geophysics 92 (2013) 50–56 53
3. Discussion
Fig. 7. Comparison of stacked sections: not deconvolved (A), with the application of TVSW (B), and with the application of pre-correlation predictive deconvolution (C). CDP interval is 5 m.
the interval from 10 to 40 Hz. In fact, the most prominent effects of the comparable to the predictive deconvolution before cross-correlation
nonunity prediction lag are suppression of high frequencies and preser- (Fig. 5B); for small offsets the pre-correlation predictive deconvolution
vation of the overall spectral shape of input data (Yilmaz, 2001). Anoth- helps to resolve the reflections, removing the internal multiples (ringing).
er important point is that spiking deconvolution doesn't produce a The stacked sections presented in Fig. 7 are obtained by applying a
zero-phase wavelet because of the presence of noise (Berkhout, simple processing consisting of a correction for the geometrical spread-
1977). The phase spectrum of TVSW (Fig. 8D) shows a phase change ing and a trace equalization. Comparison between section with and
with respect to not deconvolved data. Hence, analysis of the amplitude without pre-correlation deconvolution, highlights that the application
and phase spectra shows that, in this case, predictive deconvolution of the deconvolution (before or after the cross-correlation) increases
furnished a good result preserving the signal better than spiking and the signal/noise ratio and resolution.
TVSW. The reflections at 220 ms and 500 ms are identifiable in all three sec-
Again a test was performed by applying an inverse Wiener filter tions, but in Fig. 7C, where the section with pre-correlation predictive
using as input the sweep and the Klauder wavelet as expected output deconvolution is plotted, they are clearer and more continuous. In partic-
(Fig. 6B), but the result evidences that the noise is not removed as in ular, the reflection at about 220 ms is better recognizable between sta-
the case of the deconvolution application (Fig. 5). tions 80 and 130, where strong environmental noise caused by factories
In order to compare the result of the pre-correlation deconvolution is present. In Fig. 7C, a deeper reflection at about 800 ms (near the top
with respect to the deconvolution after the cross-correlation, a TVSW of the Miocene Molasse) is highlighted. Between 0 ms and 200 ms, reflec-
was applied. In Fig. 6A, we can see how the TVSW at large offset is tions in data with pre-correlation and post-correlation deconvolution
L. Baradello, F. Accaino / Journal of Applied Geophysics 92 (2013) 50–56 55
Fig. 8. Amplitude and phase spectra: correlated data (A), correlated data with application of spiking deconvolution before cross-correlation (B), correlated data with application of
predictive deconvolution before cross-correlation (C), correlated data and application of TVSW (D).
(Fig. 7C and B) show a comparable resolution, while in data without remove mono-frequencies as for instance of 50 Hz. In this paper we ev-
deconvolution (Fig. 7A) the shallower events between stations 80 and idence that pre-correlation predictive deconvolution is a useful tool to
140 are affected by environmental noise that creates false reflections. remove other types of noise, such as traffic and factory noise.
The stacked section, shown in this paper, if compared to seismic The best result was obtained by applying a pre-correlation predic-
data acquired in a neighboring area (Giustiniani et al., 2008) affected tive deconvolution using as gap the first zero cross of the autocorrela-
by low levels of environmental noise, allows us to identify the aqui- tion function and an operator length of 500 ms, larger than the length
fers targeted at 200 ms and 500 ms. Only the shallower reflections of the Klauder wavelet (300 ms). Predictive deconvolution can be ap-
are less continuous with respect to the data of Giustiniani et al. plied to uncorrelated field data, which can be considered to be ap-
(2008) due to the greater environmental noise. Shallow structures proximately at minimum-phase. In fact, the earth acts on the sweep
evidenced in proximity of the factories are probably due to noise. In as a minimum-phase filter, so that performing the deconvolution on
fact after the application of deconvolution they are attenuated (see uncorrelated data makes the minimum phase assumption more real-
cdp 70–80 in Fig. 7). istic with respect to the cross-correlated data that are close to the
zero phase. The results evidenced that the choice of pre-correlation
4. Conclusions spiking or predictive deconvolution is a function of geometry acquisi-
tion. In our case until offset about 500 m spiking deconvolution fur-
Correlation with pilot sweep can degrade the vibroseis data collect- nishes a better resolution with respect to predictive deconvolution,
ed in areas where strong environment noise is present, causing a while for offsets greater than 500 m the last one gives better results.
spreading of the noises along the entire trace length especially in
high resolution prospecting and a short record length (in our case 1 s Acknowledgments
TWT). Many authors have performed several types of deconvolutions
to improve the signal/noise of vibroseis data. Deconvolution in vibroseis This work was supported by P.R.I.N 2008 research project
data is usually applied after cross-correlation or using zero-phase “Geophysical characterization of deep aquifer of the Friuli Venezia
deconvolution or to design a filter that converts Klauder wavelet to its Giulia western plains”.
minimum phase equivalent (Ristow and Jurczyk, 1975). Ristow and The authors are very grateful to the anonymous reviewers for
Jurczyk also applied the pre-correlation deconvolution in order to their incisive and constructive comments which improved the paper.
56 L. Baradello, F. Accaino / Journal of Applied Geophysics 92 (2013) 50–56
References Lebedev, A.V., Beresnev, I.A., 2004. Nonlinear distortion of signals radiated by Vibroseis
sources. Geophysics 69, 968–977.
Berkhout, A.J., 1977. Least-squares inverse filtering and wavelet deconvolution. Geo- Lerwill, W.E., 1981. The amplitude and phase response of a seismic vibrator. Geophysical
physics 42 (7), 1369–1383. Prospecting 29, 503–528.
Brittle, K.F., Lines, L.R., Dey, A.K., 2001. Vibroseis deconvolution: a comparison of cross- Li, X.P., Soellner, W., Hubral, P., 1995. Elimination of harmonic distortion in Vibroseis
correlation and frequency-domain sweep deconvolution. Geophysical Prospecting data. Geophysics 60 (2), 503–516.
49, 675–689. Nicolich, R., Della Vedova, B., Giustiniani, M., Fantoni, R., 2004. Carta del sottosuolo
Cambois, G., 2000. Zero-phasing the zero-phase source. Leading Edge 19, 72–75. della Pianura Friulana (Map of Subsurface Structures of the Friuli Plain). Litografia
Carlini, A., Storer, P., 1991. Seismic vibrator data processing for stratigraphic interpretation: Cartografica Firenze.
a combined deterministic–statistical approach. First Break 9, 458–466. Porsani, M.J., Ursin, B., 1998. Mixed-phase deconvolution. Geophysics 63, 637–647.
Costain, J.K., Coruh, C., Çoruh, Cahit, 2004. Basic Theory of Reflection Seismology: With Pritchett, W.C., 1990. Acquiring Better Seismic Data. Chapman and Hall Ltd., London.
Mathematica Notebooks and Examples. Elsevier, p. 571. Ristow, D., Jurczyk, D., 1975. Vibroseis deconvolution. Geophysical Prospecting 23,
Dong, L., Margrave, G., Mewhort, L., 2004. Examining the Phase Property of the 363–379.
Nonstationary Vibroseis Wavelet: SEG Expanded Abstracts, 23, p. 1961. http:// Robinson, E.A., Treitel, S., 1977. The spectral function of a layered system and the deter-
dx.doi.org/10.1190/1.1851180. mination of the waveforms at depth. Geophysical Prospecting 25, 434–459.
Abd El-Aal, A.E.-A.K., 2011. Harmonic by harmonic removal technique for improving Sallas, J.J., 1984. Seismic vibrator control and the downgoing P-wave. Geophysics 49,
vibroseis data quality. Geophysical Prospecting 59 (2), 279–294. 732–740.
Futterman, W.I., 1962. Dispersive body waves. Journal of Geophysical Research 69, Sallas, J.J., Weber, R.M., 1982. Comments on “The amplitude and phase response of a
5279–5291. seismic vibrator” by W. E. Lerwill. Geophysical Prospecting 30, 935–938.
Geyer, R.L., 1989. Vibroseis. Society of Exploration Geophysicists, Tulsa, Okla. ISBN: Tygel, M., Huck, H., Hubral, P., 1991. Mixed-delay wavelet deconvolution of the point
0931830907. source seismogram. Geophysics 56, 1405–1411.
Gibson, B., Larner, K., 1984. Predictive deconvolution and the zero phase source. Geophysics Walker, D., 1995. Harmonic resonance structure and chaotic dynamics in the earth-
49, 379–397. vibrator system. Geophysical Prospecting 43, 487–507.
Giustiniani, M., Accaino, F., Picotti, S., Tinivella, U., 2008. Characterization of shallow Yilmaz, O., 2001. Seismic Data Analysis, vol. I. Society of Exploration Geophysicists,
aquifers by high-resolution seismic data. Geophysical Prospecting 56, 655–666. Tulsa, Okla (IG. No 10).
Giustiniani, M., Accaino, F., Picotti, S., Tinivella, U., 2009. 3D seismic data for shallow
aquifers characterization. Journal of Applied Geophysics 68, 394–403.