To be successful in today’s E&P environments, petroleum geoscientists must detect, and properly image, increasingly complex reservoirs, by resolving the fine detail of ever-smaller hydrocarbon accumulations. High-quality seismic data play a key role in this task and are of great significance in the effort to reduce overall E&P risk. The demands placed on modern seismic data are multi-fold. The data must enable the identification and delineation of leads and prospects, based on pre-stack seismic, and must quantify key reservoir properties, to increase the probability of successfully separating lithology-fluid facies. All of these goals must be achieved in 3D, using all the dimensions of the seismic data, mainly pre-stack 3D, and, later on, 4D (time-lapse).
The dream of reservoir geoscientists, seeking to optimize the locations of costly wells, is to be able to discriminate the physical properties of rock formations before they are actually drilled.
Until now, seismic images have fallen short of delivering on this goal. To date, most of the seismic interpretation and reservoir characterization work performed has had to rely on using relatively “narrow” bandwidth seismic datasets, or seismic data that had a spectrum preferentially skewed toward the low or high end of the frequency range.
In 2007, PGS launched a new, dual-sensor, streamer acquisition system—developed with the objective of providing broader seismic bandwidth, without any compromise in pre-stack data quality or acquisition efficiency (Tenghamn, 2007). Results obtained over the intervening six years have demonstrated the benefits of this acquisition system for recording superior seismic data.
The main benefits of broadband seismic data relate to the increase in resolution on offer from the wider bandwidth
(Fig. 1), in addition to the improved penetration facilitated by signals that are richer in lower frequencies. This type of broadband seismic also has demonstrated improvements in interpretability.
 |
| Fig. 1. Left to right: schematic relationship of seismic velocity, well low-frequency model and seismic bandwidth. The represented amplitude spectra are in logarithmic scale, not normal scale; this is a more accurate representation of the seismic wave propagation into the earth. |
|
The arguments for broadband are so compelling that there is good reason to believe that, before long, all new marine seismic acquisition will be broadband. Various examples (Ozdemir, 2009; Reiser, 2012; Lafet, 2012; Whaley, 2013) have demonstrated the benefits of broader bandwidth, focusing mainly on the improvements in interpretability of the seismic data. What is, perhaps, not yet fully realized are applications aside from pure imaging, where broadband data have a significant impact on the outcome of geoscientific analysis, in the E&P environment.
The ability to derive physical rock properties from seismic data is of great value to geoscientists. After all, it is rocks that operators are drilling, not acoustic signals. Seismic inversion and rock physics analysis are the most commonly used techniques to derive rock or elastic properties from seismic data. A fundamental bias, of all inversion methods, is that a particular seismic data set can lead to a number of inversion results. With conventional or band-limited seismic data, there is a lack of information at the high and low ends of the amplitude spectrum.
Geoscientists must, therefore, constrain the number of possible solutions to reach the most reliable one. Normally, this is done by injecting known, or “a priori,” information into the model, using nearby well data or other geological values. As a consequence, the uncertainty of the results away from these constraints increases significantly, Fig. 2. Using broadband seismic data, with its greater frequency content, substantially reduces the amount, and potential bias, of a priori data input. It, therefore, makes the inversion or any quantitative interpretation solution less dependent on what we already believe, and increases its usefulness in areas where a priori information may be scarce or uncertain.
 |
| Fig. 2. The process of estimating reservoir properties from the well location will be more reliable and predictable, if the seismic bandwidth is broader. |
|
CASE STUDIES
Having laid the groundwork, and described the rationale behind broadband seismic, the potential benefits will now be illustrated. The case studies are subdivided into specific aspects, such as the estimation of elastic properties, with and without well information.
Elastic properties derivation without well information. The first case study is in the northern part of Brazil’s Santos basin—on trend and/or adjacent to some prolific discoveries, including the Lula, Carioca, and Guara pre-salt reservoirs.
This survey was acquired in late 2012, with the main focus being on the sub-salt reservoirs. Over this dataset, PGS performed time and pre-stack depth processing, aimed at enabling clearer imaging focused on the pre-salt interval. A significant effort was made to build an accurate velocity model for the depth migration; the broader bandwidth of the dual-sensor helped in this regard. When pre-stack analysis is compared with the adjacent, modern, conventional seismic data set, with deep sub-salt targets, the uplift is evident in terms of:
- Improved penetration and imaging of deep targets, especially at the pre-salt level
- Better signal-to-noise ratio, especially of deeper sub-salt targets
- Broader seismic frequency bandwidth, especially at the low end
- Improved fidelity of pre-stack elastic attributes (Fig. 3)
 |
| Fig. 3. Acoustic impedance, estimated using dual-sensor streamer seismic, is definitively clearer and exhibits an overall higher signal-to-noise ratio and more low frequencies (left-hand figure) than conventional streamer seismic. |
|
Figure 3, which focused deep in the section (around 5.5 to 6.5 sec), demonstrates that, at the level of sub-salt targets, the range of the low frequencies is significantly extended with the dual-sensor streamer data, as compared to the conventional data, across all offset/angles without losing any high frequencies.
At the pre-salt interval, the dual-sensor data show over an octave of additional low frequencies. It is the additional low frequencies and improved signal to noise penetrating deeper that enable better elastic attributes to be derived in the sub-salt objectives.
In this context, where well information is very sparse, and the objective is relatively deep and in a complex geological setting, having reliable seismic information would be crucial for the identification and delineation of a prospect. The more stable and precise elastic attributes derived from pre-stack, dual-sensor volumes lead to more accurate and reliable reservoir property prediction, which has significant value when seeking to de-risk a potentially costly deepwater prospect.
Elastic properties derivation without well constraint. In this case study, dual-sensor streamer seismic data are compared to conventional seismic data over Grevling field, a Jurassic oil discovery in the central North Sea. For this field, which lies in a mature basin, the main imaging challenge was that the field represents a relatively deep target, with a consolidated sandstone reservoir zone of interest below a high-impedance chalk interval. The conventional dataset was acquired in 2005, but it has been reprocessed recently through a modern processing sequence; the dual-sensor seismic was acquired in 2010. Therefore, a fair comparison can be made of the two datasets, Fig. 4.
 |
| Fig. 4. Conventional and dual-sensor frequency panels, decomposed from the seismic stack by stationary wavelet transform, demonstrate the octave of difference in low-frequency content and, in this case, the preservation of the high-frequency content. |
|
Over this field, the full suite of well logs is available. This well information has been used to build a reliable rock-physics model, showing the elastic properties behavior of rocks at this depth, in the presence of hydrocarbons and brine. These results were then compared with the elastic properties estimated, using pre-seismic inversion of the conventional and dual-sensor seismic performed, without this well information as constraints. Hence, the well information in this case acts as blind well cross-validation, to investigate if the broader bandwidth of the dual-sensor dataset would result in increased accuracy of the elastic properties prediction.
As shown in Fig. 4, the conventional seismic lacks at least one octave at the low frequency end of the amplitude spectra, at the reservoir level; the two seismic datasets are equivalent, in the frequency content, at the high end of the spectrum, due to the fact that the reservoir interval is quite deep. This results in the highest frequencies being attenuated by the time the seismic energy reaches the target zone between 2.6 sec and 2.7 sec.
Following this initial analysis of the seismic data, simultaneous pre-stack seismic inversion was performed, without well constraints, by integrating the seismic velocity at the low-frequency end, to attempt to estimate the absolute elastic properties. The results of the “absolute” simultaneous inversion, in terms of acoustic impedance and Vp/Vs ratio, were compared with the derived rock physics model, Fig. 5.
|
| Fig. 5. Left: Acoustic impedance estimated without any well data. Right: Crossplots with data highlighted, with points determined by the top reservoir and OWC from the well data. The grey points show the scatter of all points within the cube of data around the reservoir. The colored ellipses show the predicted response from an independent well model for two depths covering the reservoir interval. |
|
As can be seen in Fig. 5, the dual-sensor streamer data acoustic impedance is less noisy and characterizes the reservoir extent very well at the top of the structure. The workflow has a potential weakness in recovering large density variations, such as a thick hard chalk unit toward the top of the section, which should be characterized by a hard (blue) impedance response. This lack of response is due mainly to the simple density approximation; in this case a density model based on well data would introduce the large density variations needed to accurately estimate the chalk properties. However, at the zone of interest, and in the reservoir interval, the pre-stack simultaneous inversion matches the reservoir extent relatively well.
The cross-plots on the right-hand side of Fig. 5 display the rock properties extracted from the inversion volumes, based on polygons picked by using the well information, to accurately map the top reservoir, OWC, and extent of brine sand recorded in the wells. The dual-sensor, streamer seismic inversion results demonstrate several improvements over conventional data, such as a better fit to the predicted rock physics model (the colored ellipses at two depths); an improved background trend in the data that matches—but was not derived from—well data; and a much better discrimination between the brine and oil sand. An identical constant was added to the Vp/Vs values for both datasets; however, the dual-sensor data show a distinct Vp/Vs background shale trend recorded in the low frequency of the seismic data, whereas the conventional data mainly show variation around the constant value.
It is important to note that the rock physics model was derived independently from the available well log data; it was not used, in any way, to constrain the pre-stack inversion results. Although the match between impedance and Vp/Vs ratios—computed from the seismic data and the well-derived rock physics templates—is not perfect, Fig. 5 certainly demonstrates the superior quality of the dual-sensor results due, principally, to its richer, low-frequency content and preservation of the high frequencies.
CONCLUSIONS
The examples presented in this article demonstrate that using dual-sensor streamer technology not only has operational benefits, such as a wider weather window, due to its deeper towing capabilities, but more importantly, that the extended bandwidth of the data benefits both structural and quantitative seismic interpretation of leads, prospects and existing reservoir units.
The examples presented are just a small part of the whole story, and more will be available in the future, but they already highlight the main points: broader, recorded frequency bandwidth seismic data (incorporating both the low and high ends of the amplitude spectra) significantly increase the ability of explorers or developers to improve their confidence and accuracy of elastic and reservoir properties prediction away from well constraints.
The last example further highlighted some important aspects: a clear background shale trend can now be estimated with dual-sensor streamer seismic data, as opposed to conventional band-limited seismic. This gives additional confidence in exploration settings that it is possible to derive good approximations of reservoir properties from dual-sensor broadband data, even with no or limited well control.
The impact of this new type of seismic is across the asset life—from an exploration setting, where little or no well information exists, to more developmental settings, where this type of data can definitively optimize the field development. 
ACKNOWLEDGEMENTS
The author thanks his colleagues at PGS—who produced the technical material—for all the fruitful discussions.
REFERENCES
- Lafet, Y., L. Michel, R. Sablon, D. Russier and R. Hanumantha, “Variable depth streamer—benefits for rock property inversion,” 74th EAGE Conference & Exhibition, Extended Abstracts, Copenhagen, Denmark, 2012.
- H. Özdemir, Unbiased deterministic seismic inversion: more seismic, less model, First Break, 27(11), 43-50, 2009.
- Reiser, C., T Bird, F. Engelmark, E. Anderson and Y. Balabekov, Value of broadband seismic for interpretation, reservoir characterization and quantitative interpretation workflows, First Break, 30(9), 67-75, 2012.
- Tenghamn, R., S. Vaage, S. and C. Borresen, A dual-sensor, towed marine streamer: its viable implementation and initial results, 77th SEG Annual Meeting, Expanded Abstracts, 989–993, San Antonio, Texas, 2007.
- Whaley, M., Anderson, E. and C. Reiser, Rock property estimation using dual-sensor streamer data without well constraint: North Sea Jurassic case study, 75th EAGE Conference & Exhibition, Extended Abstracts, Houston, Texas, 2013.
|
