Sparse hyperbolic inversion of seismic data

• US 6,219,621 B1
• Filed: 06/25/1999
• Issued: 04/17/2001
• Est. Priority Date: 06/30/1998
• Status: Expired due to Fees

• Current Assignee: Exxonmobil Upstream Research Company (Exxon Mobil Corporation)
• Original Assignee: Exxonmobil Upstream Research Company (Exxon Mobil Corporation)
• Inventors: Hornbostel, Scott C.
• Primary Examiner(s): Patidar, Jay
• Assistant Examiner(s): Taylor, Victor J.
• Application Number:
  US09/344,099
• Time in Patent Office: 662 Days
• Field of Search: 702/16, 702/14, 367/72, 367/38, 324/345
• US Class Current: 702/16
• CPC Class Codes: G01V 1/36 Effecting static or dynamic corrections on records, e.g. correcting spread; Correlating seismic signals; Eliminating effects of unwanted energy
• CPC Class Codes: G01V 1/364 Seismic filtering G01V1/37 takes precedence
• CPC Class Codes: G01V 2210/56 De-ghosting; Reverberation compensation

Abstract

A method of sparse hyperbolic inversion is used to suppress multiples in marine seismic data. By explicitly including the two-way water-bottom reflection traveltime and water-bottom reflection coefficient as part of an augmented model domain, the inversion may be successfully carried out even for intermediate water depths and sloping water bottoms. Linear noises can also be suppressed by appropriate definition of the model. The invention also has the capability of handling time-varying wavelets or several different wavelets simultaneously. This latter capability is useful in suppressing dispersive ground roll on land seismic data.

Claims

1. A method for suppressing water-bottom multiples in a gather of seismic data traces, said method comprising the steps of:

(a) inverting said gather of seismic data traces from the data domain to the model domain to obtain a model domain representation of said gather, each data sample in said model domain representation being defined in terms of a set of parameters including at least zero-offset traveltime, moveout velocity, two-way water-bottom reflection traveltime, and water-bottom reflection coefficient;

(b) determining a forward modeling matrix that relates the model domain to the data domain; and

(c) using said model domain representation and said forward modeling matrix to identify and remove said water-bottom multiples from said gather of seismic data traces.

2. The method of claim 1, wherein said step of using said model domain representation and said forward modeling matrix to identify and remove said water-bottom multiples from said gather comprises forward modeling each non-zero data sample in said model domain representation of said gather to only a primary reflection hyperbola in the data domain to obtain a data domain representation of only the primary reflections in said gather.

3. The method of claim 2, wherein said forward modeling matrix contains an estimate of the seismic wavelet for said gather of seismic data traces.

4. The method of claim 3, wherein said estimate of the seismic wavelet is obtained from measurement of the actual seismic wavelet used to generate said seismic data traces.

5. The method of claim 3, wherein said estimate of the seismic wavelet is obtained from an isolated pulse in said seismic data traces.

6. The method of claim 3, wherein said estimate of the seismic wavelet is a time-varying wavelet.

7. The method of claim 3, wherein said estimate of the seismic wavelet comprises a set of wavelets.

8. The method of claim 1, wherein said step of using said model domain representation and said forward modeling matrix to identify and remove said water-bottom multiples from said gather further comprises the steps of:

(a) modifying said forward modeling matrix to model only said water-bottom multiples;

(b) forward modeling said model domain representation of said gather to the data domain using said modified forward modeling matrix to obtain a data domain representation of said water-bottom multiples; and

(c) subtracting said data domain representation of said water-bottom multiples from said gather of seismic data traces.

9. The method of claim 8, wherein said forward modeling matrix contains an estimate of the seismic wavelet for said gather of seismic data traces.

10. The method of claim 9, wherein said estimate of the seismic wavelet is obtained from measurement of the actual seismic wavelet used to generate said seismic data traces.

11. The method of claim 9, wherein said estimate of the seismic wavelet is obtained from an isolated pulse in said seismic data traces.

12. The method of claim 9, wherein said estimate of the seismic wavelet is a time-varying wavelet.

13. The method of claim 9, wherein said estimate of the seismic wavelet comprises a set of wavelets.

14. The method of claim 1, wherein said step of inverting said gather of seismic data traces from the data domain to the model domain comprises an iterative optimization procedure.

15. The method of claim 14, wherein said iterative optimization procedure comprises the steps of:

(d) selecting an initial size for said model domain representation and an initial set of input data traces to be modeled;

(e) iteratively (i) generating synthetic data corresponding to said model domain representation, (ii) comparing said synthetic data to said input data traces using a residual error objective function, and (iii) updating said model domain representation, until the error calculated according to said residual error objective function is less than a preselected maximum error;

(f) iteratively (i) increasing the number of input data traces, (ii) reducing the size of said model domain representation, (iii) updating said residual error objective function, and (iv) repeating step (e), until all of said seismic data traces have been modeled.

16. The method of claim 15, wherein said residual error objective function is updated using an L_p-norm.

17. The method of claim 1, wherein said set of parameters includes water bottom dip.

18. A method for suppressing noise in a gather of seismic data traces, said method comprising the steps of:

(a) inverting said gather of seismic data traces from the data domain to the model domain to obtain a sparse model domain representation of said gather, each data sample in said sparse model domain representation being defined in terms of a set of parameters including at least zero-offset traveltime and stacking velocity, each primary reflection in said gather being represented by a single data sample in said sparse model domain representation;

(b) performing a filtering operation on said sparse model domain representation of said gather to suppress data samples relating to said noise; and

(c) forward modeling said filtered sparse model domain representation of said gather to the data domain.

19. The method of claim 18, wherein said step of forward modeling comprises multiplying said filtered sparse model domain representation of said gather by a forward modeling matrix, said forward modeling matrix containing an estimate of the seismic wavelet for said gather.

20. The method of claim 19, wherein said forward modeling matrix includes a plurality of linear moveouts, each having an associated wavelet, to permit modeling of dispersive noises.

21. The method of claim 18, wherein said step of inverting said gather of seismic data traces from the data domain to the model domain comprises an iterative optimization procedure.

22. The method of claim 21, wherein said iterative optimization procedure comprises the steps of:

(d) selecting an initial size for said model domain representation and an initial set of input data traces to be modeled;

(e) iteratively (i) generating synthetic data corresponding to said model domain representation, (ii) comparing said synthetic data to said input data traces using a residual error objective function, and (iii) updating said model domain representation, until the error calculated according to said residual error objective function is less than a preselected maximum error;

(f) iteratively (i) increasing the number of input data traces, (ii) reducing the size of said model domain representation, (iii) updating said residual error objective function, and (iv) repeating step (e), until all of said seismic data traces have been modeled.