Relative amplitude preservation in high-resolution shallow reflection seismic: a case study from Fore-Sudetic Monocline, Poland
- 180 Downloads
The acquisition parameters and methodology of seismic data processing for high-resolution seismic imaging viewed through relative amplitude preservation are presented. An example of the obtaining of high-quality, shallow seismic data with a variable end-on spread is shown. The source used for the project is an accelerated weight drop. The study area lies within the mine waste disposal area, near Rudna village (Fore-Sudetic Monocline, WS Poland), and results are given for a 2D experimental profile. The aim of the project was to design optimal acquisition and processing parameters for the detailed recognition of Tertiary deposits. The proposed acquisition parameters are a compromise between time, cost and results. High-resolution seismic imaging enables the determining of layers within the range of thicknesses between 5 and 15 m, while the maximal depth of imaging reaches 400 m.
KeywordsHigh-resolution seismic Thin layers Acquisition Weight drop Noise attenuation Relative amplitude processing
Throughout the last 30 years, the interest in engineering and environmental applications of seismic methods for high-resolution imaging of the near surface has increased intensively. Scientists have come to realize that high-resolution imaging is key to understanding the complex nature of many geological targets better (e.g. Berkhout 1985; Sheriff 1985, 1997).
Exploration geophysics is focused on structural and reservoir depth range. It can reach up to several kilometers, registering frequency band spans comprising several tens of Hertz, although the dominant frequency is typically below 50 Hz, depending on the source parameters and the target depth. In contrast to deep seismic imaging, shallow seismic uses sources of significantly lower energies but with a much higher frequency range, guaranteeing higher resolution of data. Higher frequencies are attenuated more efficiently, however, limiting the depth range of such a signal. The application of near-surface seismic imaging is used mainly for engineering purposes (Steeples and Miller 1990), recognition of shallow geological structures, e.g., depositional sequences (Baumgart-Kotarba et al. 2001), subglacial structures (Benjumea and Teixido 2001), post-glacial deposits (Francese et al. 2007; Baumgart-Kotarba et al. 2008), fault zones (Myers et al. 1987; Treadway et al. 1988; Miller et al. 1989, 1990) and shallow exploration targets, for example zinc ore (Singh 1984), sulfur ore deposits (Dec 2012; Dec and Cichostępski 2017a, b) and coal beds (Palmer 1987).
The idea of applying high-frequency sources to shallow seismic exploration has been discussed for many years (e.g. Knapp and Steeples 1986b). The comparison of different sources can be found in many research papers, for example, Miller et al. (1986, 1992, 1994) or Atanackov and Gosar (2013), an interesting comparison of seismic sources can be found presented by Miller et al. (1986)—concerning piezoelectric sources, hammering, weight dropping, rifle shot and others actions. The signals obtained have very wide frequency bands and dominant frequencies of 120–200 Hz. The depth range presented by the authors reaches approximately 200 ms. Similar results are presented by Steeples and Miller (1990) where the depth ranges are within 200–300 ms. A particularly interesting idea was the application of a .5 caliber rifle shot and vertical stacking that enabled the reaching of high-frequency data for a depth range of up to 1000 m (Steeples and Miller 1990). Nevertheless, the explosive sources, such as rifle or sparker, requires the drilling of shallow wells (holes) that are filled with water for which permission should be issued by the local authorities. This increases the costs and extends the time needed for seismic acquisition.
In shallow seismic, weight-based sources are common and the acquisition parameters (live channels, offset range, fold) are proportionally smaller than in large-scale exploration seismic surveys. The problem of overriding importance in shallow seismic imaging is the maintaining of the widest frequency bands with a low noise content—high signal-to-noise ratio (S/N ratio). Very low fold value governs the need for adjusting an appropriate acquisition design, taking into account geometry, source and receiver parameters and processing sequences for the specific target. A low fold parameter also causes a resolution enhancement at the processing step, which must often be compensated for by a significant alteration of seismic wavelet. This enhancement is highly undesirable—every arbitral change of a seismic wavelet results in a false reservoir interpretation and should be omitted at all costs. In the case of imaging shallow exploration targets and lithofacial analysis, it is crucial not to disturb the relative amplitude ratios which are almost impossible to interpret when a seismic wavelet is distorted. Petrophysical parameter determination relies on amplitude value and its changes (Dec and Cichostępski 2017a, b; Kwietniak et al. 2018), so it is highly necessary to keep seismic with relative amplitude preservation (RAP) versions for such purposes.
RAP, originally defined as reflection amplitude preservation (Sheriff 1991), is “processing and display designed to preserve relative amplitudes of seismic reflections.” It was a Western Geophysical trademark, later to be used generally as a name for seismic volumes that underwent no amplitude changes. RAP-processed sections enable linking amplitude changes with lateral and vertical changes in petrophysical properties of a geological medium. Only such a volume can be treated as input data to any reservoir (petrophysical) analyses, such as the inversion process, AVO analysis or instantaneous attribute application. The processing criteria for relative amplitude preservation were stipulated by Shuey (1985).
RAP, altogether with the need for high-resolution data, was a key factor in our research. In comparison with the previously mentioned results (Steeples and Miller 1990), the dominant frequency is lower, around 125 Hz, but the result can still be considered as high-resolution seismic data. This was the result of an accurate choice of the sensors used and the processing algorithms. In the following paper, we would like to present the methodology that was a tradeoff between result quality and all other costs behind the seismic acquisition. Our approach can easily be applied in similar geological settings.
Currently, RAP processing is the standard approach in exploration seismology. It is rarely applied in near-surface studies. We want to show that this kind of restricted processing can be also exploited in high-resolution shallow seismic investigation. RAP-processed high-resolution shallow seismic data can be used further for the imaging of petrophysical parameters distribution, in the near-surface zone.
Location and geological setting
The geological sequence within the study area consists of Proterozoic, Carboniferous, Permian (Lower Permian—Rotliegend formation and Upper Permian—Zechstein), Triassic, Tertiary (Palaeogene and Neogene) and Quarterly (Kłapciński et al. 1984).
The Triassic is represented by the complete profile of Lower Triassic sandstones—Roethian (Olenekian and Induan). The Lowermost Triassic is built from arkosic and quartzitic sandstones; the thickness of these deposits is about 300 m. The Middle Bundsandstein (Middle Lower Triassic) is comprised of arkosic sandstones with a band of brown and dark gray shales. Dolomites and marls represent Ret with platies of limestones and anhydrite. The thickness of Ret reaches up to 140 m.
Uppermost in the profile lays Palaeogene (Eocene and Oligocene) and Neogene (Miocene and Pliocene). The Eocene consists of quartz-rich and glauconite sandstones with thicknesses of 20 m. The Oligocene is represented by similar lithologies with a small admixture of gravel of thicknesses of 20 m. Brown coal beds—the Głogowski beds—closes the Paleogene profile. The Miocene profile consists of gravels, quartz sands and green and gray argil. Within the Miocene sediments, there are three brown coal beds: Ścinawski, Łużycki and Henryk coal beds (Kłapciński and Peryt 2007). Toward the top of the profile, above the Henryk coal bed, there are argillic Pliocene sediments, which have an interlining of quartz sandstones. The thickness of the Pliocene deposits is locally less than 100 m.
The Quaternary sediments have thicknesses of several tens of meters. They consist of post-glacial deposits, mostly rich in quartz sandstones, gravels and gray clays. Within the gray clays, additives of sands and gravel beds can often be encountered.
Within the Tertiary sediments, thin bed sequences of sandstone and argil, as well as brown coal beds, are common. The tops and beds of the coal layers are very strong reflective boundaries, thus producing a rapid decrease in energy (attenuation) carried by the elastic field through the deeper intervals, causing a decreased resolution of seismic image below the coal bed sequences.
In some places, the complex glaciotectonic processes have caused the Pliocene illites to be overthrust into the Quaternary sediments also causing difficulties in structural interpretation. Adding to the fact that the Quaternary sediments are rather loose, which also results in substantial attenuation of a seismic signal, the whole geological interval under analysis requires a well thought out acquisition design and processing approach.
Field test of geophones
The 14-Hz geophones have flat frequency characteristic from 80 Hz, but maximal amplitude frequency (dominant frequency) lies within 8–30 Hz. The usage of 14-Hz geophones results in increased dynamics of low-frequency components and consequently higher detectability and visibility of surface wave. The 100-Hz geophones attenuate much more efficiently the low frequencies that cause the surface wave and other low-frequency noise, producing less dominance on the field record. For the same velocity of displacement the shape of the frequency response above 100 Hz is similar, but in bandwidth between 8 and 80 Hz, the response from 14 Hz sensors has much higher amplitudes. To verify these characteristics, the field test of two sets of sensors was conducted.
The test results show that in the case of 100-Hz geophones, amplitudes for frequencies 0–50 Hz are much attenuated and the spectrum amplitudes within a bandwidth of 50–200 Hz are highly favorable (Fig. 3b). The − 12 dB bandwidth equals 120 Hz (40–160 Hz). The − 6 dB bandwidth registered by 100-Hz geophones is between 60 and 130 Hz (Fig. 3b), and for 14 Hz it is between 30 and 80 Hz (Fig. 3a).
Frequency spectra show that the signal generated by ESS-500 Turbo has a broad frequency characteristic. Between frequencies of 100–200 Hz, the amplitude values are significant and the application of deconvolution should result in a wide and flat spectrum. According to Knapp and Steeples (1986a), 100-Hz geophones perform better in the 75–500 Hz bandwidth (Fig. 3b) and attenuate unwanted low-frequency signals. The operating characteristics of high-frequency geophones are better than those of low-frequency geophones, when recording in the same ground motion.
With the use of 100-Hz geophones, much high-frequency noise can be registered, especially when the conditions for a shot are poor (wet soil, uneven surface, etc.). To overcome this problem, vertical stacking should be performed. Five or more stacks are sufficient to obtain noise-free field recording (Fig. 3).
To obtain high-resolution data, the authors decided to use geophones of higher frequency characteristics.
Seismic acquisition parameters
ESS-500 Turbo (accelerated weight drop)
Single vertical 100-Hz geophone
Variable end-on roll-along spread
3 times average, max up to 10
Absolute offset range
The acquisition designed enabled us to obtain results of good quality with a satisfactory signal-to-noise ratio. The optimal acquisition parameters and proper offset distance, causes the coherent noise—surface wave and airwave to appear at times greater than the target time (see Fig. 3).
In such a case, amplitudes of these waves (i.e. surface and air waves) do not disturb the response from the geological target and do not have to be cut out at the processing step. Often the procedures that filter out coherent noises influence the amplitude relation that is unfavorable (Yilmaz 1987; Steeples and Miller 1998). The choice to use the end-on spread, instead of the split-spread, enables longer offsets, which is essential for reservoir properties, reasoning basing on AVO (Amplitude vs. Offset) analysis (Chopra and Castagna 2014; Dec and Cichostępski 2017a).
In the case of the profile, the shot offset was variable, and not all field records hold the criterion for coherent noise (surface and air waves) to be below the time to the target. Nevertheless, 12 traces within one CDP (common depth point) are sufficient to easily reduce the level of unwanted linear events in the stacking process. Residuals from these noises may be effectively filtered on the stacked seismic section.
Data processing aimed at relative amplitude preservation (RAP) is much more rigorous than processing path for structural interpretation purposes. The RAP processing should compensate the energy loss, reduction in coherent and non-coherent noises, reduction in the influence of low velocity layer and complete reduction in all artefacts on seismic traces caused by the processing steps themselves.
It is important not to introduce any procedures that can change the shape and amplitudes of the wavelet. The resulting volume (CDP gathers and stacks) should preserve the true reflectivity of the subsurface (true amplitude processing). It is hardly possible not to affect the reflectivity, and it is a commonly accepted practice to maintain only relative amplitudes; hence, the processing sequence is relative amplitude processing (RAP processing). RAP processing sequence should apply to all procedures to satisfy Shuey’s equations (Shuey 1985; Cambois 2001; Aki and Richards 2002; Chopra and Castagna 2014). Usually, a standard path of processing is applied (Resnick 1993; Yilmaz 2001), but some procedures must be adjusted, especially if the target lies within shallow depths (Steeples and Miller 1998). In RAP processing, the only amplitude recovery procedures that can be applied are surface-consistent scaling and spherical divergence correction. AGC, time-variant scaling, etc., do not apply in RAP processing; furthermore, 2D filters, e.g. median, FK, etc., should not be applied, since it can disturb relative amplitudes (see “Appendix”).
Seismic processing sequence
Geometry and trace edit
High-line frequency rejection (50 Hz)
Geometric divergence correction (offset dependent)
First break muting
Surface-consistent amplitude scaling (source and receiver)
Surface-consistent spike deconvolution (source and receiver)
Bandpass filtering (40/60–200/250 Hz)
Automatic residual statics
Bandpass filtering (40/60–200/250 Hz)
Signal enchantment with the preserved signal (4D-DEC)
The initial pre-processing procedures include data format change, geometry adjustments, high-line rejection of 50 Hz (due to electric lines near the study area), spherical divergence correction, refraction statics and first break muting.
The gradual loss of amplitude, resulting from geometrical spreading, is a significant effect which needs compensation. For RAP processing purposes, appropriate amplitude compensation is required. The correction is achieved by approximating the effect of amplitude loss due to waveform divergence. According to Ursin (1990), geometrical spreading manifests itself by amplitude loss with an increasing offset. For far offsets and shallow seismic boundaries, the loss is substantial. Because of this, apart from applying time correction, offset correction was also applied. With offsets of up to 134 m, it is especially valid for imaging layers within depths between 30 and 150 m.
Refraction statics were calculated employing the routine approach, from slopes and intercept times of picked first breaks. We have defined a two-layer velocity model; the estimated velocity in the first layer was 600 m/s, and in the second 1750 m/s. After static correction, we carefully removed first breaks by a top mute with the preservation of wide-angle shallow reflections.
In RAP processing, variability of reflection amplitudes in land seismic should be caused only by changes in petrophysical parameters between layers. However, variability in amplitude, phase and frequency partially depends on changes in source and receiver conditions. For example, unconsolidated material (e.g., sand or mud) at shot or receiver can significantly change the character of amplitude at that particular location compared to an adjacent position that is of a hard ground surface. As a consequence, changing near-surface conditions deteriorate repeatability of the source. Also, differences in receiver coupling can distort the seismic signal. To compensate for these effects, a surface-consistent method is used (Taner and Koehler 1981). The surface-consistent approach means that a single operator is applied for all traces that have the same surface point in common (e.g., all traces within a single shot or receiver). The operator is solved by Gauss–Seidel iterative process.
To compensate for the effects of variations in source and receiver conditions and the coupling of seismic record amplitudes, we applied surface-consistent amplitude scaling. In this process, the average amplitude of each trace is computed, and the computed trace scalar is assumed to be a multiplication of shot, receiver, offset and CMP. For relative amplitude preservation, only shot and receiver components are used.
Next, for further improving vertical resolution and balancing the amplitude spectra, a surface-consistent spike deconvolution was applied. The length of the operator was set to 50 ms. The surface-consistent approach in deconvolution compensates for near-surface wavelet distortion. The average wavelet of each trace is assumed to be a convolution of shot, receiver, offset and CMP wavelet. Subsequently, the deconvolution operator is calculated from the amplitude spectra of the chosen components. Similarly to surface-consistent amplitude scaling, for relative amplitude preservation only shot and receiver components were used (Cary and Lorentz 1993). After surface-consistent deconvolution, the 40/60–200/250 Hz bandpass filtering was applied.
In Fig. 5b, the resulting shot after the application of these procedures is presented. The spectrum is flattened within 50–200 Hz, and the dominant frequency is approximately 125 Hz. Spectrum has white character.
The 4D-DEC algorithm works by creating wavelet within small overlapping windows in time and space. In each window, a single dip is picked, and the wavelet is created by first break stacking along the chosen dip. Events are then aligned and restocked with correlation statics. With this operation, a high-resolution wavelet is produced. The obtained wavelet is matched in amplitude and time to each trace. The design window time length that we applied was 200 ms, maximum dip 8 ms/trace, filter length consisted of 13 traces and maximum static 5 ms (see “Appendix”). The algorithm preserves fine details of statics and the signal amplitude, while the trace wavelet within the window is constant. In result, 4D-DEC creates a noise model. The volume of noise computed in this way (Fig. 6b) was adaptively subtracted from the input data (Fig. 6a) without damaging the signal. The resulting stack volume is free from noise (Fig. 6c).
Results and discussion
The surface seismic source ESS-500 generates a wavelet of very high resolving power. The length of the wavelet is usually between 18 and 22 ms, depending on the specific conditions in the field. After the deconvolution process, the wavelet is much shorter and compressed which gives a broad and flat spectrum.
The final stack with interval velocities displayed in the background is shown in Fig. 8a. For display purposes, a mean amplitude scaling was applied for all the traces with a single scalar. Therefore, relative amplitudes are preserved across the section. The middle frequency for the final stack, similarly to the extracted wavelet, is about 125 Hz, which in our case gives a high resolution. For the Tertiary sediments (20–380 ms), the interval velocities around 1400–2000 m/s (Fig. 8a) for the dominant frequency of 125 Hz correspond to the wavelength (λ) of 11–16 m. With the resolution criteria given by Widess (1973), the limit of resolution for thin layers is equal to λ/4 for 2D seismic data. In the extreme case, i.e., for thin coal beds that have very low velocities (c.a. 1250 m/s) and have opposite reflection coefficients in their tops and beds, the minimal thickness that can be resolved is 2.5 m.
Beneath, in the Miocene interval two coal beds are interpreted: first the Łużycki coal bed (thickness of about 10 m) and second in depth the Ścinawski coal bed of a thickness of 5 m (which is very close to the resolution limit of 2.5 m). In Fig. 9, we also present an enlarged image of the Łużycki and Ściniawski coal beds. The thickness of the Łużycki bed is approximately equal in wavelength, which makes the resulting resolution high, whereas the thickness of the Ściniawski bed is approximately equal to half of the wavelength.
Other seismic horizons visible at section are associated with layers of argil and gravels. At a depth of about 320 m, there is a weak reflection corresponding to the top of the Głogowski coal bed (the top of a thin Eocene complex). Between depths of 340 and 360 m lies the seismic horizon which is linked to the top of the Tertiary deposits.
The seismic image presented in Fig. 9 shows how rapid the energy loss is with increasing depth, when strong seismic boundaries (coal beds) are present. Excepting structural imaging, as a result of RAP processing, we can also observe horizontal variability of a real wavelet’s amplitudes along horizons. It is most important for coal prospecting, where these changes can be correlated with facial changes within brown coal seams.
The test acquisition enabled us to verify that ESS-500 seismic generator is a suitable source for shallow seismic imaging. For Fore-Sudetic monocline’s Tertiary sediments, the ESS-500 enables high-resolution data to be obtained. The generated wavelet can be effectively compressed in the deconvolution process, the key step to maintaining high frequencies for the given dataset. The resulting wavelet is very short (40/60–200/250 Hz bandwidth, 8 ms period, 12 ms length). Well-designed seismic acquisition can significantly limit the distortion of the seismic signal at the processing step since coherent noises do not interfere with the geological target.
For the interval velocities of about 2000 m/s and with the middle frequency of 125 Hz, the dominant wavelength is 16 m. In such a case, the limit of resolution reaches at least 4 m. For brown coal beds of much lower velocities (Vp = 1250 m/s), the limit of resolution decreases to about 2.5 m—this result is exceptionally good for shallow seismic imaging.
A comparison of different denoising methods indicated that the 4D-DEC procedure is the most optimal for maintaining the criteria of relative amplitude preservation.
RAP-processed high-resolution, shallow seismic data may be the appropriate input for further precise recognition of lithofacial changes, and evaluation of petrophysical and geomechanical parameters at the near-surface zone.
The research was supported by the Polish Academy of Sciences via European Plate Observing System (EPOS) project and AGH-UST University of Science and Technology in Krakow, Poland grant no. 184.108.40.2065. Vista 2D/3D Seismic Data Processing was provided by the Schlumberger through University Software Grant Program. We would like to thank the editor M. Malinowski and two anonymous reviewers for providing valuable and insightful feedback on the manuscript.
Compliance with ethical standards
Conflict of interest
On behalf of all authors, the corresponding author states that there is no conflict of interest.
- Aki K, Richards PG (2002) Quantitative seismology. W.H. Freeman and Company, New YorkGoogle Scholar
- Baumgart-Kotarba M, Dec J, Ślusarczyk R (2001) Quaternary tectonic grabens of Wróblówka and Piniążkowice and their relation to Neogene strata of the Orava Basin and Pliocene sediments of Domański Wierch series in Podhale, Polish West Carpathians. Studia Geomorfolo Carpatho-Balcanica XXXV:101–119Google Scholar
- Baumgart-Kotarba M, Dec J, Kotarba A, Ślusarczyk R (2008) Glacial trough and sediments infill of the Biała Woda Valley (the High Tatra mountains) using geophysical and geomorphological methods. Studia Geomorphol Carpatho-Balcanica XLII:75–108Google Scholar
- Berkhout AJ (1985) Seismic resolution: a key to detailed geologic information. World Oil 201:47–51Google Scholar
- Butler P (2012) White noise suppression in the time domain. CSEG Rec 37(9):39–44Google Scholar
- Cambois G (2001) AVO processing: myths and reality. CSEG Rec 26:30–33Google Scholar
- Chopra S, Castagna JP (2014) AVO. Investigations in geophysics. Society of Exploration Geophysics, TulsaGoogle Scholar
- Dec J (2012) High-resolution seismic survey for recognition of the Osiek sulphur deposits and determination of dynamic changes resulting from exploitation. Wydawnictwa AGH, Kraków [in Polish with English abstract] Google Scholar
- Dec J, Cichostępski K (2017a) Evaluation of sulphur deposit properties on the basis of geomechanical parameters. Zeszyty Naukowe Instytutu Gospodarki Surowcami Mineralnymi i Energią PAN 101:203–216 [in Polish with English abstract] Google Scholar
- Dec J, Cichostępski K (2017b) Estimation of sulphur deposit resources on the basis of seismic data. Zeszyty Naukowe Instytutu Gospodarki Surowcami Mineralnymi i Energią PAN 101:217–227 [in Polish with English abstract] Google Scholar
- Francese RG, Hajnal Z, Schmitt D, Zaja A (2007) High resolution seismic reflection imaging of complex stratigraphic features in shallow aquifers. Memorie Descriptive della Carta Geologica d’Italia LXXVI:175–192Google Scholar
- Gisco (2018) Documentation of Gisco vertical geophones. http://www.giscogeo.com/products-equipment/geophones. Accessed 31 July 2018
- Kłapciński J, Peryt TM (2007) Budowa geologiczna monokliny przedsudeckiej. Monografia KGHM Polska Miedź SA 69–77Google Scholar
- Kłapciński J, Konstantynowicz E, Salski W, Kenig E, Preidl M, Dubińsdki K, Drozdowski S (1984) Atlas obszaru miedzionośnego (monoklina przedsudecka). Wydawnictwo Śląsk, KatowiceGoogle Scholar
- Pożaryski W (1979) Mapa geologiczna Polski I krajów ościennych 1:1000000. Wyd. Geologiczne, WarszawaGoogle Scholar
- Resnick JR (1993) Seismic data processing for AVO and AVA analysis. In: Castagna JP, Backus MM (eds) Offset-dependent reflectivity—theory and practice of AVO analysis. Society of Exploration Geophysics, Tulsa, pp 175–189Google Scholar
- Sheriff RE (1985) Aspects of seismic resolution. In: Berg OR, Woolvetron DG (eds) Seismic stratigraphy II: an integrated approach to hydrocarbon exploration. AAPG Memoir 39Google Scholar
- Sheriff RE (1991) Encyclopedic dictionary of exploration geophysics. Society of Exploration Geophysics, Tulsa, p 240Google Scholar
- Sheriff RE (1997) Seismic resolution: a key element. AAPG Explor Geophys Corner 18:44–51Google Scholar
- Steeples DW, Miller RD (1990) Seismic reflection methods applied to engineering, environmental, and groundwater problems. In: Ward S (ed) Review and tutorial: investigations in geophysics, vol 5. Society of Exploration Geophysicists, Tulsa, pp 1–30Google Scholar
- Yilmaz O (1987) Seismic data processing. Society of Exploration Geophysics, TulsaGoogle Scholar
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.