10.11.2012 Views

Wave Equation PSDM - Petroleum Geo-Services

Wave Equation PSDM - Petroleum Geo-Services

Wave Equation PSDM - Petroleum Geo-Services

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

A Publication of PGS <strong>Geo</strong>physical Vol. 5 No. 9<br />

October 2005<br />

<strong>Wave</strong> <strong>Equation</strong> Pre-Stack Depth Migration at PGS<br />

Introduction<br />

Pre-stack depth migration (<strong>PSDM</strong>)<br />

of 3D seismic data is a challenging and<br />

complex process. An almost countless<br />

array of <strong>PSDM</strong> solutions can be found<br />

in the literature. Each solution will<br />

pursue an improvement in a specific<br />

geophysical component of imaging<br />

(e.g. the accuracy of steep dips or the<br />

preservation of “true” amplitudes), or<br />

will attempt to overcome the significant<br />

computational costs that have delayed<br />

the entry of <strong>PSDM</strong> into mainstream<br />

processing. As computational<br />

resources become faster and cheaper it<br />

becomes more feasible to use <strong>PSDM</strong> on<br />

larger size 3D data volumes.<br />

Simultaneously, it becomes more<br />

feasible to implement increasingly<br />

sophisticated <strong>PSDM</strong> solutions, thereby<br />

addressing challenges to successful<br />

seismic imaging from increasingly<br />

complex geology.<br />

The industry market for<br />

commercial <strong>PSDM</strong> processing has<br />

matured rapidly during the last decade.<br />

<strong>PSDM</strong> algorithms using Kirchhoff<br />

integral techniques are now routinely<br />

used on large-scale 3D volumes.<br />

Significantly, “wave equation” <strong>PSDM</strong><br />

(WE<strong>PSDM</strong>) algorithms are now being<br />

used on large 3D surveys in the Gulf of<br />

Mexico, and smaller 3D volumes<br />

elsewhere.<br />

WE<strong>PSDM</strong> algorithms use<br />

wavefield extrapolation techniques to<br />

image pre-stack seismic data in depth.<br />

In contrast to most Kirchhoff<br />

implementations, they can naturally<br />

handle “multi-arrivals” or<br />

“triplications”, where multi-pathing<br />

occurs between specific source and<br />

receiver locations. Complex scattering,<br />

focusing and defocusing effects are<br />

implicitly also accounted for.<br />

WE<strong>PSDM</strong> does not make the high<br />

frequency assumptions of Kirchhoffbased<br />

<strong>PSDM</strong> and, therefore, can<br />

produce higher quality images in areas<br />

of complex geology (rapid lateral<br />

variations in velocity). As noted, the<br />

increasingly attractive price<br />

/performance ratios of Linux-based PC<br />

clusters allow the pursuit of more<br />

accurate and more computationally<br />

intensive WE<strong>PSDM</strong> algorithms.<br />

Discussion is given below to the<br />

selection of different WE<strong>PSDM</strong><br />

algorithms for different geological<br />

scenarios.<br />

It is clear that ongoing<br />

developments in WE<strong>PSDM</strong> technology<br />

will see its application to increasingly<br />

larger projects, in a cost effective<br />

manner. From a technical viewpoint,<br />

we will eventually see wavefield<br />

extrapolators being developed that are<br />

capable of handling wave propagation<br />

at all dips in a medium with all arbitrary<br />

lateral velocity variations.<br />

<strong>Wave</strong> <strong>Equation</strong> <strong>PSDM</strong><br />

Fundamentals<br />

The many names given to WE<br />

migration methods are often confusing<br />

Continued on next page<br />

Summary<br />

PGS has invested significant<br />

resources in the development of<br />

wave equation pre-stack depth<br />

migration (WE<strong>PSDM</strong>) routines.<br />

These include Common Azimuth<br />

migration and shot migrations at<br />

increasing levels of sophistication:<br />

Gazdag Phase Shift, Split-Step<br />

Fourier, and Finite Difference<br />

methods. Gazdag Phase Shift and<br />

Split-Step Fourier are highly<br />

efficient extrapolation methods<br />

suitable for mild lateral velocity<br />

variations. Finite Difference<br />

methods are more accurate for<br />

strong velocity changes, and may<br />

be classified into explicit and<br />

implicit approaches. Each<br />

technique presents inherent merits<br />

and challenges with respect to<br />

efficiency and cost. The critical<br />

virtue is that a flexible range of<br />

WE<strong>PSDM</strong> solutions are available<br />

to be customized to the demands<br />

of any imaging project. In other<br />

words, the solution is matched to<br />

the complexity of the earth model<br />

and the geophysical objectives.<br />

When 3D acquisition is<br />

appropriately parameterized to<br />

enable precise 3D data<br />

regularization, and an accurate 3D<br />

velocity model can be constructed,<br />

WE<strong>PSDM</strong> will provide the most<br />

sophisticated and accurate<br />

imaging solutions available.


TechLink October 2005 Page 2<br />

<strong>Wave</strong> <strong>Equation</strong> <strong>PSDM</strong><br />

Fundamentals<br />

Continued from Page 1<br />

because they may refer to the domain in<br />

which the process occurs (e.g.<br />

“frequency-wavenumber”), the manner<br />

in which the data is organized during<br />

the process (e.g. “shot record” or “shotgeophone”),<br />

the methodology used (e.g.<br />

“survey sinking” or “reverse time”) or<br />

the type of extrapolation used (e.g.<br />

“explicit finite difference”).<br />

Irrespective of the many types of<br />

WE<strong>PSDM</strong> implementations, all<br />

migrations consist of the following two<br />

elements:<br />

• The solution of a differential equation<br />

(the wave equation or its equivalent),<br />

and<br />

• The application of an imaging<br />

condition.<br />

In terms of both accuracy and cost,<br />

the extrapolator used is the most<br />

defining issue. The most accurate<br />

extrapolator is that used by reverse time<br />

migration, which is a “two way”<br />

solution. Two way extrapolation can<br />

handle multiples, is able to image high<br />

dip and turning waves, provides better<br />

(relative) amplitudes preservation and<br />

better accommodates numerical noise<br />

and imaging artefacts. Unfortunately,<br />

two way extrapolators are about an<br />

order of magnitude slower than one<br />

way extrapolators, so all current<br />

commercial WE<strong>PSDM</strong> algorithms use<br />

one way (downward) extrapolators.<br />

One way extrapolators cannot handle<br />

multiples (they must be removed prior<br />

to imaging), are restricted to dips less<br />

than 90°, and will accommodate lateral<br />

velocity variations according to their<br />

complexity (and therefore cost). In<br />

order of increasing complexity,<br />

examples include Gazdag phase shift,<br />

Split-Step Fourier, phase shift plus<br />

interpolation, and (implicit or explicit)<br />

finite difference.<br />

The two most<br />

common one way<br />

extrapolation<br />

implementations are<br />

“common shot” and<br />

“survey sinking”<br />

implementations.<br />

Common shot domain<br />

techniques extrapolate<br />

the shot and receiver<br />

w a v e f i e l d s<br />

independently, and<br />

obtain an image for<br />

each shot by comparing<br />

the two wavefields at<br />

each spatial location<br />

and depth. Plane wave<br />

techniques (e.g.<br />

“delayed shot”) are a cost-saving The PGS <strong>Wave</strong> <strong>Equation</strong><br />

implementation of common shot <strong>PSDM</strong> Portfolio<br />

techniques that migrate several shots PGS currently offers five<br />

simultaneously. Survey sinking WE<strong>PSDM</strong> solutions contained in three<br />

techniques (such as common or narrow routines in its Cube Manager<br />

azimuth WE<strong>PSDM</strong>) simultaneously<br />

extrapolate both the shot and receiver<br />

wavefields, effectively simulating the<br />

wavefield as if it was recorded at each<br />

extrapolated depth (thus the “survey<br />

sinking” name). The output image at<br />

each spatial location is obtained when<br />

the extrapolated shot and receiver are<br />

collocated (i.e. zero offset and time).<br />

Common shot implementations are the<br />

most accurate WE<strong>PSDM</strong> solutions, and<br />

form the majority of the PGS<br />

WE<strong>PSDM</strong> portfolio.<br />

Numerous wavefield extrapolation<br />

techniques are available within the PGS<br />

WE<strong>PSDM</strong> portfolio. The choice of<br />

which to use is driven by: 1. The<br />

complexity of the velocity model,<br />

notably the lateral variability, and 2.<br />

The computational speed requirements<br />

of the particular project. The benefits<br />

of a large portfolio of WE<strong>PSDM</strong><br />

solutions are flexibility and efficiency,<br />

as illustrated in Figure 1.<br />

TM<br />

Figure 1: The variety of wavefield extrapolators available<br />

within the PGS WE<strong>PSDM</strong> portfolio. Gazdag phase shift<br />

migration is exact for no lateral velocity variation, and is<br />

the fastest available option (much faster than Kirchhoff).<br />

Finite difference migration provides the most accurate<br />

solution in complex velocity media (including VTI<br />

anisotropy), but is also the most computationally expensive.<br />

PGS does provide the ability to select the most efficient<br />

extrapolation solution at each depth step, constrained by<br />

the lateral variability in the velocity model at the output<br />

depth step.<br />

processing system:<br />

1. A fast Common Azimuth WE<strong>PSDM</strong><br />

that is particularly suitable for typical<br />

marine seismic geometries, and for<br />

early iterations in depth imaging<br />

when the velocity model is being<br />

refined. Lateral velocity variations<br />

are elegantly handled, as is the phase<br />

correction required at each depth step.<br />

2. A unique, multi-algorithm Adaptive<br />

WE <strong>PSDM</strong> that efficiently images<br />

geology where the overburden has<br />

relatively mild velocity variation.<br />

The strategy involves the automated<br />

selection of the most appropriate<br />

algorithm based on the measured<br />

velocity complexity at each depth<br />

step (refer also to Figure 1). Thus,<br />

Gazdag Phase Shift and Split-Step<br />

Fourier methods are used where mild<br />

velocity variation allows. An<br />

inline/cross-line splitting Implicit<br />

Finite Difference solution, with phase


Page 3 A Publication of PGS <strong>Geo</strong>physical<br />

correction, is applied in the case of<br />

more complex velocity variation. Our<br />

adaptive approach is also available in<br />

shot and plane wave modes, and<br />

incorporates WKBJ corrections in the<br />

wavefield extrapolation for correct<br />

amplitude preservation. The<br />

application may also be run entirely in<br />

single-algorithm mode for each of the<br />

three respective algorithms.<br />

3. An Explicit Finite Difference<br />

WE<strong>PSDM</strong>. This is the PGS flagship<br />

WE<strong>PSDM</strong> implementation, and uses<br />

an innovative “constrained” operator<br />

design for downward extrapolation<br />

that ensures stability, efficiency and<br />

high dip accuracy. The<br />

implementation handles isotropic and<br />

VTI anisotropic velocity fields, steep<br />

dips, and is available in both full shot<br />

and plane wave modes. Efficiency<br />

measures inherent in the<br />

implementation include the ability to<br />

handle rectangular bins and the<br />

adaptive design of variable step sizes.<br />

Of significance to the extraction of<br />

reservoir parameters, this solution<br />

includes an imaging condition that<br />

extracts data reflectivity.<br />

Naturally, the accuracy of the 3D<br />

velocity model used for WE<strong>PSDM</strong> is<br />

critical, and represents a critical<br />

challenge to any imaging success. PGS<br />

has the ability to generate (reflection)<br />

angle gathers for velocity model<br />

updates, and the ability to perform<br />

velocity scanning for velocity model<br />

updates, particularly in sub-salt areas.<br />

Figure 2 illustrates an explicit<br />

finite difference WE<strong>PSDM</strong> application<br />

to the well-known SIGSBEE2A<br />

synthetic model.<br />

Figure 3 illustrates the<br />

improvements possible by replacing<br />

Kirchhoff <strong>PSDM</strong> with explicit finite<br />

difference WE<strong>PSDM</strong>, as applied to a<br />

Continued on next page<br />

Figure 2: The explicit finite difference <strong>PSDM</strong> enables very accurate imaging below salt<br />

structures, as illustrated with this SIGSBEE2A example. Subtle faults and sub-salt structures<br />

are evident, and image focus is excellent in all spatial locations. Thus, WE<strong>PSDM</strong> is proven to<br />

have significant value in locations of complex velocity variation.<br />

Figure 3: Kirchhoff <strong>PSDM</strong> (left) vs. Explicit Finite Difference WE<strong>PSDM</strong> (right) in the Gulf of<br />

Mexico. Note the benefits of WE<strong>PSDM</strong> for imaging sedimentary features on the flanks of the<br />

salt and below the salt.<br />

Figure 4: Isotropic (left) vs. Anisotropic (right) Explicit Finite Difference <strong>PSDM</strong> comparison<br />

of images from the Marmousi model dataset. Incorporation of VTI anisotropy parameters<br />

allows superior focusing of events, notably in the upper half of the section.


TechLink October 2005 Page 4<br />

The PGS <strong>Wave</strong> <strong>Equation</strong><br />

<strong>PSDM</strong> Portfolio<br />

Continued from Page 3<br />

real dataset is from the Gulf of Mexico.<br />

Significant improvements were<br />

achieved in terms of image resolution<br />

and clarity.<br />

Finally, Figure 4 illustrates the<br />

benefits of VTI anisotropic capability<br />

when the explicit finite difference<br />

WE<strong>PSDM</strong> is applied to the well-known<br />

Marmousi synthetic model.<br />

Optimum Acquisition<br />

Considerations for WE<strong>PSDM</strong><br />

All “wave equation” processes<br />

assume that the input dataset is<br />

uniformly sampled. As any 3D<br />

streamer acquisition survey will<br />

inherently involve variable vessel<br />

trajectories, streamer feathering, and<br />

variable sea state, pre-stack 3D data is<br />

by nature irregularly sampled. Aside<br />

from lower computational cost,<br />

Kirchhoff migration methods have been<br />

the historical method of choice because<br />

they are far more robust in their ability<br />

to accommodate irregular data. The<br />

solution to this dilemma is effective<br />

data “regularization”. 3D<br />

regularization is a significant area of<br />

C O N T A C T<br />

PGS <strong>Geo</strong>physical<br />

London<br />

Tel: 44-1932-260001<br />

Fax: 44-1932-266465<br />

Oslo<br />

Tel: 47-67-52-6400<br />

Fax: 47-67-52-6464<br />

Houston<br />

Tel: 1-281-509-8000<br />

Fax: 1-281-509-8500<br />

Singapore<br />

Tel: 65-6735-6411<br />

Fax: 65-6735-6413<br />

© 2005 <strong>Petroleum</strong> <strong>Geo</strong>-<strong>Services</strong>. All Rights Reserved<br />

ongoing research within the industry<br />

and academia, and has a history as<br />

challenging as that of the development<br />

of imaging methods themselves.<br />

Whilst outside the scope of this<br />

discussion, it can be briefly observed<br />

that regularization methods have a full<br />

spectrum of technical complexity vs.<br />

cost efficiency. Compromises in the<br />

integrity of regularization processing<br />

will translate to compromises in the<br />

integrity of the final imaging result. In<br />

the most sophisticated scenario, Fourier<br />

regularization is not theoretically<br />

limited to plane events or a limited<br />

number of dips, does not use a<br />

geophysical model to describe the input<br />

data, and provided that the input data is<br />

not unacceptably aliased, can be<br />

considered as an “exact” process.<br />

Therein lays an assumption with<br />

implications for the ideal 3D<br />

acquisition platform. High Density 3D<br />

(HD3D) acquisition presents the<br />

acquisition platform that minimizes<br />

data aliasing in any direction, and can<br />

be customized to satisfy specific<br />

assumptions about cross-line aperture<br />

or target illumination, as demanded by<br />

other high-end processing algorithms.<br />

When allowed to work successfully,<br />

Fourier regularization will use a<br />

parametric inversion to output the data<br />

to a perfectly uniform data grid, as<br />

demanded by the user. Properly<br />

sampled (HD3D) data will present the<br />

best platform for regularization, which<br />

in turn will present the optimum<br />

platform for the successful application<br />

of WE<strong>PSDM</strong>.<br />

Figure 5: Dense spatial sampling of the reflected wavefield is provided by HD3D acquisition<br />

to avoid aliasing and ensure the successful application of regularization and WE<strong>PSDM</strong><br />

technology. The “dark” regions on the wavefront shown (from a single shot in a complex<br />

model, as produced by 3D ray tracing) are “multi-arrivals”. A specific receiver location at<br />

the surface will receive three discrete arrivals from the same wavefront. This observation<br />

reinforces the value of dense sampling for accurate reconstruction of the target during<br />

(WE<strong>PSDM</strong>) imaging.<br />

For Updates on PGS Technological Advances, visit www.pgs.com<br />

More TechLinks at www.pgs.com/techlink

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!