Communications – electrical: acoustic wave systems and devices – Seismic prospecting – Land-reflection type
Reexamination Certificate
2003-04-30
2004-09-28
Moskowitz, Nelson (Department: 3663)
Communications, electrical: acoustic wave systems and devices
Seismic prospecting
Land-reflection type
C367S051000, C702S017000
Reexamination Certificate
active
06798714
ABSTRACT:
FIELD OF THE INVENTION
The present invention relates to processing seismic data, and is particularly concerned with avoiding normal moveout stretch during the stages of normal moveout correction and common-midpoint stacking.
BACKGROUND OF THE INVENTION
Referring to
FIG. 1
there is illustrated a known setup
10
for gathering seismic exploration data. A seismic sound source
12
is generated at or just below the earth's surface
14
, or in the case of marine seismic, just below the water surface. For each such source generation, or “shot”, the sound travels into the earth
16
, reflects off of changes in geology (called “events” or “reflectors”)
18
, and travels back to the surface
20
, where it is simultaneously recorded at a plurality of receivers
22
. A single such recording at a receiver
22
, is called a “trace” and is in the form of a regularly sampled time series measuring the particle velocity (for land data) or change in acoustic pressure (for marine data). A single shot is typically recorded at hundreds or thousands of receivers simultaneously. Many such shots are taken for a single seismic data set, sometimes resulting in hundreds of millions of seismic traces.
Seismic data can be used to interpret, or infer, geology, and thus is useful for the location, identification, and exploitation of petroleum and minerals. Before it can be used for this purpose, however, seismic data must undergo a series of statistical processes, a task generally referred to as “seismic processing”.
There are a number of effects that can confound the ability to interpret seismic data. One such effect is noise, defined generally as any unwanted recorded energy. The origins of noise can be both natural and man-made, much of the noise being caused by the shot itself.
A second confounding effect is multiples, which is energy that has reflected more than once in its propagation from shot to receiver. A typical multiple propagation path is illustrated in FIG.
2
. Seismic energy travels from the shot
12
down
24
to a reflector
26
, back up
28
to the surface
14
, back down
32
to a second reflector
18
, and then back up
34
a receiver
36
where it is recorded.
The desired energy for seismic exploration is singly reflected, that is energy that has reflected from only one geological reflector. Events that have reflected only once before being recorded, are referred to as “primaries”.
One step in seismic processing is normal-move-out (NMO) correction. Traces with different “offsets” (the horizontal distance between shot and receiver) have the same reflection (or event) appearing at different times. NMO correction is a time-variant shifting of sample values so that each trace's reflections are aligned to occur at the same time, that is, as if the trace had zero offset. The NMO correction allows for stacking of traces with different offsets.
The principal parameter involved in applying NMO correction is a stacking velocity, which is a single value, varying in both time and space, controlling the amount of time shift. Choosing these stacking velocities is a routine part of seismic processing that occurs for virtually every data set.
For a given zero-offset time within a single common midpoint (CMP) gather, a traditional NMO correction shift follows a hyperbolic curve as a function of offset. In recent years more complicated formulas have been introduced that compensate for near-surface effects (Link, et al, 1992), anisotropy (change in velocity with propagation direction), and vertical velocity gradient. To correct for these last two effects, the seismic processor must pick a NMO correction parameter &eegr; (Alkhalifah, 1998) in addition to a stacking velocity.
Another step in seismic processing is front-end muting, which is the setting to zero of sample values near the beginning of the trace. The purpose of front-end muting is to remove noise and other unwanted effects from the front of the trace.
Another step in seismic processing is common-midpoint (CMP) stacking, where traces, having roughly the same midpoints between their shot and receiver positions, are collected into groups. At each recorded time sample, the non-muted values for every trace in the group are averaged together, producing a single “stacked” trace for each group. One benefit of CMP stacking is noise reduction due to the averaging of many values into one value. A second benefit is multiple reduction, resulting from the fact that while NMO correction lines up primary reflections, it does not line up multiple reflections, so that multiples tend to attenuate during averaging.
The multiple-reducing property of stacking depends critically on there being a broad range of offsets within the non-muted sample values. Another-benefit of stacking is reducing the amount of data, typically by a factor between 10 and 100, so that the data can be displayed in a manner that is convenient and easy to interpret, specifically as a “stacked CMP section”. Yet another benefit of stacking is reduction of the computation time required for later processes such as migration and noise reduction.
The typical processing steps of seismic processing are illustrated in FIG.
3
. We begin with a CMP gather of traces
42
. A NMO correction is applied
44
, and then front-end muting
46
. Finally the traces are stacked into a single stacked trace
48
.
NMO correction causes distortion of seismic events in time (Dunk and Levin, 1973), the principal effect being stretch, which is the conversion of high frequencies into low frequencies by expanding the time base. It is a well known principle in seismic processing that to be as interpretable as possible, seismic events should be as broad band in frequency as the noise allows (Berkhout, 1984). Thus NMO stretch can mean a loss in the ability to infer geology. NMO stretch is particularly severe at early times, large offsets, and fast vertical changes in velocity. Referring to
FIG. 4
, there is illustrated an artificial CMP gather
50
. After NMO correction
52
the gather has stretch
54
at far offsets and early times. After stacking the gather
56
, there is distortion of early events
58
as compared to later events
60
.
NMO stretch is caused by the implicit assumption that seismic events occur instantaneously. However, this is not the case. A seismic event typically has an effective length between 20 to 60 ms in duration. As a result, during standard NMO correction, a different time shift is applied to the beginning of an event than to the end.
A well known way of avoiding NMO stretch is to apply a front-end mute that zeroes all trace samples suffering from too much NMO stretch.
FIG. 5
illustrates a is known seismic processing sequence. The same artificial NMO-corrected gather
52
as in
FIG. 4
is shown. A front-end mute is applied
64
. The resulting stack
66
shows much less distortion at early times
68
as compared to later times
70
.
However, there are drawbacks to this approach. First, the CMP gather has less redundancy (or “fold”) at early times, resulting in decreased noise reduction in stacking. Second, the CMP gather has less far-offset information at early times, resulting in decreased multiple reduction in stacking. Third, changes in event character with offset contains valuable interpretive information referred to as amplitude-versus-offset, or AVO effects (Castagna and Backus,1993). Front-end muting results in the loss of some AVO information.
Other solutions have been suggested for mitigating NMO stretch. Rupert and Chun (1975) introduced block move sum NMO, where traces are subdivided into overlapping blocks of samples. Each block has constant-shift NMO applied, and the blocks summed with weights to form the NMO-corrected gather. A related approach was described by Shatilo and Aminzadeh (2000), where the normal moveout function is kept constant in the vicinity of discrete events. Byun and Nelan (1997) apply time-varying filters to NMO-corrected traces to reverse the loss of high frequencies.
Hicks (2001) describes a method for removing NMO stretch during stacking based on
Kelman Technologies Inc.
Moskowitz Nelson
LandOfFree
Method of performing stretch-free normal moveout (NMO) and... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Method of performing stretch-free normal moveout (NMO) and..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Method of performing stretch-free normal moveout (NMO) and... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3234735