Shoudong Huo

Elimination of land internal multiples based on the inverse scattering series

Despite the explosion of new, innovative technologies in the area of multiple identification and subsequent attenuation, their applicability is mostly limited to marine environments especially in deep water. In land seismic data sets however, the application of such multiple-elimination methodologies is not always straightforward and in many cases poor results are obtained. The unique characteristics of land seismic data (i.e., noise, statics and coupling) are major obstacles in multiple estimation and subsequent elimination. The well-defined surface multiples present in marine data are rarely identifiable in land data. Particularly in desert terrains with a complex near surface and low-relief structures, surface multiples hardly exist. In most cases, we are dealing with so called “near-surface-related multiples.” These are primarily internal multiples generated within the complex near surface.

The inverse scattering series approach towards the elimination of land internal multiples

The estimation and subsequent elimination of internal multiples in land seismic data is one of the most challenging steps in data processing. Although marine multiple elimination techniques, such as the SRME technology, are well established, in land their implementation is not straightforward and in many cases poor results are obtained. In this paper we use theoretical concepts from the Inverse Scattering Series (ISS) formulation and develop computer algorithms for land internal multiple elimination. The key characteristic of the ISS-based methods is that they do not require any information about the subsurface, i.e., they are fully data driven. Internal multiples from all possible generators are computed and adaptively subtracted from the input data. These methodologies can be applied preand post-stack and their performance is demonstrated using realistic synthetic and field datasets from the Arabian Peninsula. These are the first published results of the application of the ISS internal multiple attenuation method to the daunting challenge of land internal multiples.

Improving adaptive subtraction in seismic multiple attenuation

In seismic multiple attenuation, once the multiple models have been built, the effectiveness of the processing depends onthesubtractionstep.Usuallytheprimaryenergyispartially attenuated during the adaptive subtraction if an L 2 -norm matching filter is used to solve a least-squares problem. The expandedmultichannelmatchingEMCMfiltergenerallyis effective, but conservative parameters adopted to preserve theprimarycouldleadtosomeremainingmultiples.Wehave managed to improve the multiple attenuation result through aniterativeapplicationoftheEMCMfiltertoaccumulatethe effect of subtraction. A Butterworth-type masking filter based on the multiple model can be used to preserve most of the primary energy prior to subtraction, and then subtraction canbeperformedontheremainingparttobettersuppressthe multiples without affecting the primaries. Meanwhile, subtraction can be performed according to the orders of the multiples, as a single subtraction window usually covers different-order multiples with different amplitudes. Theoretical analyses, and synthetic and real seismic data set demonstrations,provedthatacombinationofthesethreestrategiesiseffective in improving the adaptive subtraction during seismic multipleattenuation.

Simultaneous sources deblending via frequency-varying filter in cross-spread azimuth-offset domain

Simultaneous sources acquisition, also referred to as “blended acquisition”, involves recording two or more shots simultaneously. It allows for denser spatial sampling and can greatly speed up the field data acquisition. Thus, it has potential advantage to improve seismic data quality and reduce acquisition cost. In order to achieve the goal of blended acquisition, a deblending procedure is necessary. It attenuates the interference and thus improves the resolution of the pre-stack time migration image. In this paper, we propose an efficient deblending method, which applies frequency-varying median and mean filters to cross-spread azimuth-offset gathers (XSPR-AO). The method can be used with variable window sizes according to the characteristics of the interference. The effectiveness of the method is validated by a field data example.

Complex Semblance and Its Application

Semblance, a measure of multi-trace coherence, has been used extensively in seismic data processing and interpretation such as velocity analysis and fault detection. The traditional algorithm has a difficulty at zero-crossings of seismic recordings. This problem is alleviated by applying a smoothing window at the cost of losing vertical resolutions. In this paper, we improve the algorithm by computing semblance from complex traces. Our initial results show that the complex semblance is smooth at zero-crossings. Because the smoothing time window becomes unnecessary, the higher vertical resolution can be achieved by using small windows or none. Some geological features, like faults and unconformities, appear clearer and easier to identify with the complex semblance. As the advantages are obvious and the implementation is straight-forward with the Hilbert transform, this new algorithm may replace the traditional one in future applications.

A Synthetic Simultaneous Source Wide-azimuth Acquisitions Study over a Complex Marine Environment

Marine blended source acquisition is becoming increasingly important in the seismic industry due to the possibility of reducing costs through higher productivity and improving seismic data quality through denser source sampling. One major drawback of marine blended acquisition is the crosstalk noise generated by the nearly simultaneous firing of the source arrays. It is essential to understand the characteristics of this type of noise interference and identify proper acquisition techniques and processing workflows to reduce its effects on image quality. We start by 3D finite difference modelling of a complex subsurface. We then combine the synthetic shots to simulate a four boat, wide azimuth (WAZ) marine seismic survey comprised of two streamer vessels with sources, and two additional source vessels located between the streamer spreads and off the tails of the streamers. All four sources fire nearly simultaneously with a randomized time lag of up to 500 milliseconds between sources. Overall, the gain from near-simultaneous source firing versus a conventional four-vessel WAZ marine design is an increase of approximately 2.67 times in terms of both source density and fold. This type of blended acquisition survey can also be acquired in about the same amount of acquisition time as a conventional four vessel WAZ survey. In general, most of the processing techniques for the simultaneous source blended data rely on the fact that crosstalk noise exhibits coherency in the shot domain, but appears random when viewed in a different data domain such as common channel, offset and midpoint domains. A number of processing techniques are applied in order to optimally deblend the data. These techniques are tested in several different domains and also in a cascaded manner. The results and effectiveness of each technique is evaluated and compared against the original non-blended synthetic data. Direct comparisons between the processed blended data and the single source non-blended data reveal comparable seismic images in both the prestack and post-stack domains.

Seismic Simultaneous Sources in Land: Challenges and Opportunities

essing of this optimally designed seismic blended dataset without applying any deblending algorithm produces satisfactory results. To improve the prestack analysis such as first break picking, noise removal and velocity analysis, deblending methodologies and workflows were also applied. In this paper, we present processing results related to land simultaneous sources acquisition. Novel deblending schemes will be shown along with their effectiveness in a production environment. Statics, surface consistency, noise attenuation and velocity estimation are seen as the main challenges for the processing of land blended data. Current processing schemes rely mostly on deblending so that conventional workflows can be subsequently employed for data analysis. Full blended data processing in land is still an open issue. Current practice dictates that we have to have well in advance nearsurface and velocity macromodels in order to proceed. Migration will do the rest. The real question therefore is: a) Do we develop tools to fully process blended data? or b) Do we develop effective deblending algorithms & proceed in a conventional manner? We all know the answer…it can be found in the middle!!! Of course everything starts with acquisition…do we play the game safe and acquire optimally distance-separated data or go wild and acquire data in a random fashion? Processing now becomes the key. In short, seismic acquisition and processing must be considered simultaneously!!