Technical Field

This invention relates to the general subject of seismic exploration and, in particular, to methods for identifying structural and stratigraphic features in three dimensions.

Background Of The Invention Ordinary 2-D seismic data is acquired along lines (See lines 10 and 11 in FIG. 1 ) that consist of geophone arrays onshore or hydrophone streamer traverses offshore. Geophones and hydrophones act as sensors to receive energy that is transmitted into the ground and reflected back to the surface from subsurface rock interfaces 12. Energy is usually provided onshore by vibroseis vehicles which transmit pulses by shaking the ground at pre-determined intervals and frequencies on the surface. Offshore, airgun sources are usually used. Subtle changes in the energy returned to surface often reflect variations in the stratigraphic, structural and fluid contents of the reservoirs.

In 3-D seismic the principle is similar, however, lines and arrays are more closely spaced (See FIG's. 1 and 2) to provide more detailed sub¬ surface coverage. With this high density coverage, extremely large volumes of digital data need to be recorded, stored and processed before final interpretation can be made. Processing requires extensive computer resources and complex software to enhance the signal received from the subsurface and to mute accompanying noise which masks the signal.

Once the data is processed, geophysical staff compile and interpret the 3-D seismic information in the form of a 3-D cube (See FIG. 4) which effectively represents a display of subsurface features. Using the data cube, information can be displayed in various forms. Horizontal time slice maps can be made at selected depths (See FIG. 5). Using a computer workstation an interpreter can slice through the field to investigate reservoir issues at different horizons. Vertical slices or sections can also be made in any direction using seismic or well data. Time maps can be converted to depth to provide a structural interpretation at a specific level. Three-dimensional (3-D) seismic is being used extensively worldwide to provide a more detailed structural and stratigraphic image of subsurface reservoirs. Acceptance of 3-D seismic has accelerated during the last five

years based on a proven track record that continues to grow. The 3-D payout has been measured by increased reserve estimates, cost savings from more accurate positioning of delineation and development wells, improved reservoir characterization leading to better simulation models, and the ability to predict more accurately future opportunities and problems during the production history of a field. More importantly, 3-D seismic has also been used as an exploration tool to reduce drilling risk in structurally complex areas and to predict reservoir quality in undrilled areas.

As good as 3-D seismic surveys and interpreters have become, improvements are needed.

In particular, seismic data has been traditionally acquired and processed for the purpose of imaging seismic reflections. Changes in stratigraphy are often difficult to detect on traditional seismic displays due to the limited amount of information that stratigraphic features present in a cross-section view. Although such views provide an opportunity to see a much larger portion of these features, it is difficult to identify fault surfaces within a 3-D volume where no fault reflections have been recorded. More importantly, seismic data is not known to have been acquired or used for the purpose of imaging seismic discontinuities instead of seismic reflections.

Summary Of The Invention In accordance with the present invention, a method is disclosed for the exploration of hydrocarbons. The method comprises the steps of: obtaining a set of seismic signal traces distributed over a pre-determined three-dimensional volume of the earth; dividing the three-dimensional volume into a plurality of vertically stacked and generally spaced apart horizontal slices; dividing each of the slices into a plurality of cells that are arranged into laterally extending rows and columns and that have portions of at least three generally vertically extending seismic traces located therein; measuring across each of the cells the cross-correlation between one pair of traces lying in one vertical plane to obtain an in-line value and measuring the cross-correlation between another pair of traces lying in another vertical plane to obtain a cross-line value that are estimates of the time dip in an in¬ line direction and in a cross-line direction; combining the in-line value and the cross-line value to obtain one coherency value for each of the cells; and displaying the coherency values of the cells across at least one horizontal slice.

This technique is particularly well suited for interpreting fault planes within a 3-D seismic volume and for detecting subtle stratigraphic features in 3-D. This is because seismic traces cut by a fault line generally have a different seismic character than traces on either side of the fault. Measuring trace similarity, (i.e., coherence or 3-D continuity) along a time slice reveals lineaments of low coherence along these fault lines. Such coherency values can reveal critical subsurface details that are not readily apparent on traditional seismic sections. Also by calculating coherence along a series of time slices, these fault lineaments identify fault planes or surfaces. Numerous other advantages and features of the present invention will become readily apparent from the following detailed description of the invention, the embodiments described therein, from the claims, and from the accompanying drawings.

Brief Description of the Drawings

FIG. 1 illustrates an arrangement of geophones to obtain 3-D seismic data from the earth's subsurface for processing in accordance with the present invention;

FIG. 2 is a plan view of the arrangement shown in FIG. 1 ; FIG. 3 is a representation of the seismic traces laying in a plane passing through one row of geophones shown in FIG. 2;

FIG. 4 is a pictorial representation of the information obtained from processing 3-D seismic data;

FIG. 5 is a pictorial representation of a horizontal time slice of 3-D seismic data processed in accordance with the prior art; and

FIG. 6 is a pictorial representation of a horizontal time slice of 3-D seismic data processed in accordance with the present invention.

Detailed Description While this invention is susceptible of embodiment in many different forms, there is shown in the drawings, and will herein be described in detail, one specific embodiment of the invention. It should be understood, however, that the present disclosure is to be considered an exemplification of the principles of the invention and is not intended to limit the invention to the specific embodiment or algorithm so described.

The first step is to obtain a set of seismic data in the form of seismic signal traces distributed over a three dimensional volume of the earth.

Methods by which such data is obtained and reduced to digital form for processing as 3-D seismic data are well known to those skilled in the art.

The next step is to generate a "discontinuity cube." This is done by applying a coherency algorithm to the 3-D seismic data. This algorithm may take many forms. Whatever its form, its function is to compare the similarity of nearby regions of seismic data within the 3-D seismic volume. If a trace segment is similar to its neighbors (e.g., in the in-line and cross-line directions), it is assigned a low discontinuity value; if a trace segment is not similar to its neighbors, it is assigned a high discontinuity value. FIG. 2 is a plan view of a portion of 3-D seismic volume. In order to measure discontinuity, a trace segment at one point A is compared to adjacent trace segments B and C. One way to compute trace similarity is described below.

The zero mean lagged cross-correlation in the in-line (x-direction) between trace u(t, x, y) and u(t, x+dx, y) with a lag time of "tlag" msec is defined as: k = +w u(t+k, x, y) u(t+k+tlag, x+dx, y)

Va(t, x, y) a(t, x+dx, y)

where: k = +w a(t,x,y) = ∑ u ² (t+k, x, y) k = -w

and

k = +w a(t, x+dx, y) = ∑ u ² (t+k, x+dx, y) k ■ -w

are autocorrelations used to normalize the cross-correlation, and where w+w is the length in msec of the correlation window. It is important to choose w large enough so the assumption of zero mean is valid. Values on the order of a seismic wavelet are appropriate. Other methods of normalization may be used (e.g., product of the energies of the traces, etc.). In particular, cross correlation is one method of combining two waveforms to

measure the similarities of the waveforms. Autocorrelation is a method of combining a waveform with itself. See Sheriff's "Encyclopedic Dictionary of

Exploration Geophysics," Society of Exploration Geophysicists, Tulsa,

Oklahoma.

The zero mean lagged cross-correlation in the cross-line (y-direction) between trace u(t, x, y) and u(t, x, y+dy) with a lag time of tlag msec is defined as: k = +w ∑ u(t+k, x, y) u(t+k+tlag, x, y+dy)

P _x (t. tlag) - ≡-=w

Va(t, x, y) a(t, x, y+dy)

where k =- +w a(t, x, y+dy) = ∑ u(t+k, x, y+dy) k ^' = -

The direction of apparent time dip in the x and y directions is estimated to be that lag (i.e., tlagx and tlagy) that has the greatest (i.e., most positive) cross-correlation. These values are p _x (t, tlagx) and p _y (t, tlagy).

Given the apparent dips (in msec/trace), it is a simple (but not necessarily accurate when dealing with noisy data) calculation to obtain dip and dip azimuth. More importantly, the concept of cross-correlation is extended to two dimensions by taking the geometric mean between the classical one dimensional cross-correlations:

p _xy (t, tlagx, tlagy) = Px _. t. tlagx) p _y (t, tlagy)

This value (or attribute) serves as a rather robust estimate of signal discontinuity within geologic formations as well as signal discontinuities across faults and erosional unconformities.

Computer Program A simplified FORTRAN 77 program for performing these calculations is given below:

Given a trace "x" from a 3-D seismic amplitude volume, and its two neighboring traces "y" (in the in-line direction) and "z" (in the cross-line

direction), subroutine COH calculates an output trace "rho" containing coherence coefficients using a running window cross-correlation algorithm where:

"mins" and "maxs" are the minimum and maximum sample indices for all four traces;

"inwinl" is the window length in samples;

"nlags" specifies the number of lags (relative time shifts) to do each side of "0" in the cross-correlation; and

"sr" is the sample interval in ms.

At each sample, subroutine CROSS calculates a series of normalized cross- correlation coefficients, returning the largest coefficients for each direction in "rhol " and "rho2". The time shift at which the maximum coefficients occur is returned in "tshfl" and "tshf2"; these times are not used. Subroutine COH is called repeatedly, once for every trace in the input seismic amplitude volume, to produce a new 3-D data volume or "coherency cube" containing coherence coefficients.

subroutine coh (x, y, z, rho, mins, maxs, iwinl, nlags, sr) real x(mins:maxs), y(mins:maxs), z(minstmaxs) real rho(mins:maxs)

ihwin = iwinl/2

do j = mins+ihwin, maxs-ihwin k = j - ihwin call cross (x(k), iwinl, y(k), iwinl, nlags, sr, rhol , tshfl ) call cross (x(k), iwinl, z(k), iwinl, nlags, sr, rho2, tshf2) rho(J) = sqrt (rho1 ^* rho2) enddo

return end

subroutine cross (x, nx, y, ny, lags, sr, peak, tshift) real x(0:nx-1 ), y(0:ny-1 ), sr, peak, tshift parameter (maxlags=128)

real g(-maxlags:+maxlags) double precision xx.yy

nlags = max(0, min(lags, maxlags)) tshift = 0.0 peak = 0.0 xx = 0.0 yy = 0.0 ks = 0

do ix = 0, nx-1 xx = x(ix)**2+xx enddo

if (xx .eq. 0.0) return

do iy = 0, ny-1 yy = y(iy)"2 + yy enddo

if (yy .eq. 0.0) return

do is = -nlags, + nlags g(is) = 0.0 do it = 0, nx-1 if (it-is .ge. 0) then if (it-is .le. ny-1 ) then g(is) = g(is) + x(it)*y(it-is) endif endif enddo if (abs(peak) .It. abs(g(is))) then peak = g(is) ks = is endif enddo

tshift = ks*sr

peak = peak/sqrt (xx*yy)

return

end

Landmark and GeoQuest interpretive workstations, for example, can be used to view and interpret faults and stratigraphic features by loading the discontinuity cube as a seismic volume. Visualization software (e.g.,

Landmarks's SeisCube software) may be employed to rapidly slice through the discontinuity volume to aid in understanding complex fault relationships.

Discontinuity displays can reduce interpretation cycle time when used in selecting which seismic lines to interpret, enabling the interpreter to work around faults and poo ^' r data areas. In addition, subtle stratigraphic features and complex faulting which are not readily apparent on traditional seismic displays can be rapidly identified and interpreted. FIG's. 5 and 6 are side by side comparisons of the same seismic information displayed and processed conventionally and in accordance with the present invention. Fault lines are readily apparent in FIG. 6.

Coherency maps have been run on several 3-D surveys. At depths of reasonable data quality, approximately 90% of the faults can be readily identified. Faults were identified on coherency maps which were very subtle on seismic sections, but clearly present on the coherency maps because of the robustness of the method and the map perspective of fault patterns. Since coherency maps can be run on uninterpreted time slices, the present invention offers a means to greatly accelerate mapping of the structural framework and to reveal details of fault relationships which would otherwise be interpreted only through tedious fault picking.

Specific Examples

2-D seismic coherence maps were generated along picked horizons and clearly identified shale diapirs in offshore Nigeria. In offshore Gulf of Mexico, the technique readily identified diapiric structures.

On several coherence time slices, remarkable detail of stratigraphic features, such as abandoned river channels, mud

flows, and submarine canyons, was displayed. On seismic sections, these features were sometimes apparent but, in some cases, were unidentifiable even with close scrutiny. This is the first known method of revealing fault planes within a 3-D volume where no fault reflections have been recorded. Faults are often critical to the accumulation of oil. A fault may form a seal by cutting off a structural or stratigraphic feature so the oil is trapped against the fault. On the other hand, if the fault plane contains rubble that has not been cemented, it may form a conduit for fluids. This may allow the hydrocarbons to drift up the fault plane into the feature and be trapped in it or to escape from the feature by drifting up the fault plane out of it.

Thus, fault lines can predict flow patterns in a reservoir and communication between injector and producing wells, for example. Seismic discontinuities can also provide the needed link to enable reservoir prediction between the wells and establish reservoir continuity and flow patterns across a field.

Coherency mapping of 3-D seismic is an extremely powerful and efficient tool for mapping both structure and stratigraphy. The new method is particularly sensitive to any lateral variation in wavelet character and therefore is particularly sensitive to the common causes of lateral variations in the wavelet (i.e., fault displacement or stratigraphic variations). This 3-D method analyzes a time-slice or horizon based interval and measures the maximum of the normalized cross-correlation in the in-line and cross-line directions.

Further Analysis

The discontinuity cube will clearly highlight fault planes as zones of high discontinuity. However, these zones may not be clearly in areas of lower signal-to-noise ratio. A method of enhancing these fault zones involves the application of a "median planar operator." Faults in the earth's subsurface generally express themselves as planes or surfaces. In the case of a curved fault surface, a series of small flat planes may be used to approximate the fault surface. In accordance with this aspect of the invention, a small planar operator is used to enhance (i.e., a "filter") the identification of subtle stratigraphic features. First, a small region of seismic data is selected around a center value. This region may be formed from a plurality of the cells used to form the "coherency cube." A small fault plane is then mathematically inserted into the region, and the median value of the

points within the plane is calculated for the dip and azimuth that best aligns with the zone of high discontinuity. This median value is then assigned to the center value of a new array. Next, the region of data is shifted (e.g., by one row) and the process is repeated until each point in the previously determined discontinuity cube has been analyzed as a center value. The end result is a completely new discontinuity cube with fault planes enhanced and noise and stratigraphic features (i.e., non-planar features) attenuated. These stratigraphic features may be separated by subtracting the new discontinuity cube from the old discontinuity cube without the planar filter application.

From the foregoing description, it will be observed that numerous variations, alternatives and modifications will be apparent to those skilled in the art. Accordingly, this description is to be construed as illustrative only and is for the purpose of teaching those skilled in the art the manner of carrying out the invention. Other algorithms may be used to measure the similarity of nearly regions of seismic data or to generate the "discontinuity cube." Moreover, equivalent computations may be substituted for those illustrated and described. Also certain features of the invention may be used independently of other features of the invention.

For example, stratigraphic features have been generally identified on time slices where dips were low; and consequently, the time window captured a narrow statigraphic section. In areas of higher dip, the method should work on picked horizons. Therefore, as a stratigraphic mapping tool, there is good reason to believe that new levels of detail can be mapped than previously, although this may require mapping of the horizon of interest.

As another example, while coherence slice maps by themselves are very powerful mapping tools, when used in conjunction with reconnaissance mapping of amplitudes and dip maps, there is promise of a technological milestone in mapping effectiveness for the Gulf of Mexico or similar basins with 3-D seismic. It is believed that detailed mapping of structure and stratigraphy will be accelerated by mapping in a map view and less by traditional line by line picking. Interpretation in a map view of "reconnaissance" data offers significant improvement in quality and quantity of interpretation.

Thus it will be appreciated that various modifications, alternatives, variations, and changes may be made without departing from the spirit and scope of the invention as defined in the appended claims. It is, of course,

intended to cover by the appended claims all such modifications involved within the scope of the claims.