Bayesian sequential Gaussian simulation of lithology with non-linear data

• US 5,539,704 A
• Filed: 06/23/1995
• Issued: 07/23/1996
• Est. Priority Date: 06/23/1995
• Status: Expired due to Term

• Current Assignee: Western Atlas International Incorporated (Northrop Grumman Corporation)
• Original Assignee: Western Atlas International Incorporated (Northrop Grumman Corporation)
• Inventors: Doyen, Philippe M., Den Boer, Lennert D.
• Primary Examiner(s): Moskowitz, Nelson
• Application Number:
  US08/494,603
• Time in Patent Office: 396 Days
• Field of Search: 367/14, 367/15, 367/37, 367/73, 364/428, 364/421, 364/806, 382/51
• US Class Current: 367/73
• CPC Class Codes: G01V 1/282 Application of seismic models, synthetic seismograms
• CPC Class Codes: G01V 1/48 Processing data
• CPC Class Codes: G01V 2210/665 using geostatistical modeling

Abstract

A multivariate stochastic simulation application that involves the mapping of a primary variable from a combination for sparse primary data and more densely sampled secondary data. The method is applicable when the relationship between the simulated primary variable and one or more secondary variables is non-linear. The method employs a Bayesian updating rule to build a local posterior distribution for the primary variable at each simulated location. The posterior distribution is the product of a Gaussian kernel function obtained by simple kriging of the primary variable and a secondary probability function obtained directly from a scatter diagram between primary and secondary variables.

Claims

1. A method for modeling the spatial distribution of a continuous primary petrophysical variable over a regular array of a large number of pixels representative of secondary observation stations, comprising:

a) assembling in a programmed computer including a data processor, a first set of measurements of a primary variable at a small number of control pixels and a second set of measurements of an associated secondary variable at all pixels of said array;

b) estimating the mean and variance of the Gaussian prior distribution of the primary variable from the first measurement set;

c) estimating the joint probability distribution of the primary and secondary variables from a scatter diagram of the first and second measurement sets at the control pixels;

d) selecting an as-yet unsimulated pixel at random;

e) estimating the Gaussian conditional probability distribution of the primary variable at said selected pixel by kriging the primary data at control pixels within a selected search region encompassing said selected pixel;

f) in the data processor, calculating the probability ratio between said conditional probability distribution and the prior distribution at said selected pixel;

g) extracting a one-dimensional slice through the joint probability distribution corresponding to a measured value of the secondary data at said selected pixel;

h) calculating, in the data processor, a posterior probability distribution for the primary variable at said selected pixel by taking the product of the probability ratio and the extracted one-dimensional slice;

i) drawing a simulated value for said primary variable at said selected pixel by sampling at random from said posterior probability distribution;

j) entering the simulated value of said primary variable into said array at said selected pixel as an additional control pixel;

k) repeating steps d) through j) until the primary variable is simulated for all pixels of the array;

l) with the aid of a computer graphics program, mapping the distribution of the simulated primary petrophysical variable over the array of pixels.

2. The method as defined by claim 1, comprising:

repeating steps d) through l) n times to provide n equally likely models of the simulated primary variable consistent with the secondary measurements and the known primary data measurements at said control pixels; and

averaging the n equally likely models of the simulated primary variable; and

executing step l) using the average model.

3. The method as defined by claim 2, wherein:

simulated values at each pixel are obtained from a vector including more than one secondary variable.

4. The method as defined by claim 3, wherein:

the step of estimating the Gaussian conditional distribution of the primary variable at each said selected pixel includes the step of specifying a spatial covariance model to characterize the spatial continuity of the primary variable.

11. An article of manufacture, comprising:

a global display of a model of the spatial distribution of a simulated continuous primary petrophysical variable over a region of interest produced by the process as defined by claims 2 or 9.

12. An article of manufacture, comprising:

a tangible machine-readable medium that embodies a set of instructions for programming a computer to execute the process steps defined by claim 4.

5. A method for simulating a model of the distribution of a continuous primary variable over a regular array of a plurality of pixels, comprising:

a) measuring known values of a primary variable at selected control pixels as primary control data;

b) measuring values of a secondary variable at all of said plurality of pixels;

c) estimating the mean and variance of the Gaussian prior distribution of the primary control data;

d) estimating the joint probability distribution of the primary control data and collocated secondary measurements from a scattergram of the primary control data and corresponding measurements of the secondary variable;

e) selecting an as-yet unsimulated pixel at random;

f) establishing a search region encompassing said randomly selected pixel;

g) estimating the Gaussian conditional probability distribution of the primary variable at said selected pixel by kriging the primary control data at desired control pixels within said search region;

h) calculating the probability ratio between said conditional probability distribution and said prior distribution at said selected pixel;

i) extracting a one-dimensional slice through the joint probability distribution corresponding to the measured secondary data at said selected pixel;

j) calculating a posterior probability distribution for the primary variable at said selected pixel by taking the product of said probability ratio and said extracted one-dimensional slice;

k) drawing a simulated value for said primary variable at said selected pixel by sampling at random from said posterior probability distribution;

l) entering the simulated value of said primary variable into said array at said selected pixel as an additional control pixel;

m) repeating steps e) through l) until values for the primary variable are simulated for all pixels of the array;

n) mapping the simulated values of said primary variable over said array of pixels.

6. A method for mapping the estimated spatial distribution of a continuous primary variable over an area of interest, comprising:

a) distributing a regular array of spaced-apart seismic stations over said area;

b) at selected control stations collecting actual measurements of a primary petrophysical variable as primary control data and concomitantly collecting measurements of an associated secondary seismic variable at all stations;

c) estimating the mean and variance of the Gaussian prior distribution of the primary control data;

d) calibrating measurements of the secondary variable, by estimating the joint probability distribution of the primary and collocated secondary measurements from a scattergram of the primary control data and corresponding measurements of the secondary variable;

e) selecting an as-yet unsimulated station at random;

f) defining a search region around said selected station;

g) estimating the Gaussian conditional probability distribution of the primary variable at said selected station by kriging the primary control data at desired control stations within said search region;

h) calculating the probability ratio between said conditional probability distribution and said prior distribution at said selected station;

i) extracting a one-dimensional slice through the joint probability distribution corresponding to the measured secondary data at said selected station;

j) calculating a posterior probability distribution for the primary variable at said selected station by taking the product of said probability ratio and said extracted one-dimensional slice;

k) drawing a simulated value for said primary variable at said selected station by sampling at random from said posterior probability distribution;

l) entering the simulated value of said primary variable into said array at said selected station as an additional control station;

m) repeating steps d) through

     1) until values for the primary variable are simulated for all stations of the array to define an estimate of the global distribution of the primary variable over the array of stations;

n) with the aid of a computer graphics program, displaying the global distribution.

7. The method as defined by claim 6, comprising:

executing steps d) through m) n times to provide n global distributions;

averaging the n global distributions; and

displaying the average of the n global distributions.

8. A method for mapping the estimated distribution of a continuous primary variable over an area of interest, comprising:

a) defining a regular array of spaced-apart measurement stations over said area;

b) at selected control measurement stations, collecting actual measurements of a primary petrophysical variable as primary control data and concomitantly collecting measurements of an associated secondary geophysical variable at all measurement stations;

c) estimating the mean and variance of the Gaussian prior distribution of the primary control data;

d) calibrating measurements of the secondary variable by estimating the joint probability distribution of the primary and collocated secondary measurements from a scattergram of the primary control data and corresponding measurements of the secondary variable;

e) selecting an as-yet unsimulated measurement station at random, encompassing said selected measurement station by a selected search region and estimating the Gaussian conditional probability distribution of the primary variable by kriging the primary control data within said search region;

f) calculating the probability ratio between said conditional probability distribution and said prior distribution;

g) extracting a one-dimensional slice through the joint probability distribution corresponding to the measured secondary data at said selected measurement station;

h) multiplying said probability ratio by said extracted one-dimensional slice to provide a posterior probability distribution;

i) drawing a random sample from the posterior probability distribution to provide a simulated value for the primary variable;

j) entering the simulated value of said primary variable into said array at said selected measurement station as an additional control measurement station;

k) repeating steps e) through j) until values for the primary variable are simulated for all measurement stations of the array to define the global distribution of the primary variable over the array of measurement stations;

l) with the aid of a computer graphics program, displaying the global distribution.

9. A method for modeling the spatial distribution of a continuous primary petrophysical variable over an area of interest, comprising:

a) distributing a regular array of spaced-apart pixels over said area;

b) assembling in a programmed computer including a data processor, a first set of measurements of a primary variable at a small number of control pixels and a second set of measurements of an associated secondary variable at all pixels of said array;

c) normalizing the first and second measurement sets such that the primary and secondary variables have zero means and unit variances;

d) estimating the coefficient of correlation between the primary and secondary variables from a scatter diagram of the first and second measurement sets at the control pixels;

e) selecting an as-yet unsimulated pixel at random;

f) estimating the mean and variance of the Gaussian conditional distribution of the primary variable at the selected pixel by kriging the primary data at control pixels within a designated search region encompassing said randomly-selected pixel;

g) in said data processor, calculating the mean and variance of the Gaussian posterior distribution for the primary variable at said selected pixel from the coefficient of correlation estimated in step d), the measured secondary data at said selected pixel and the mean and variance calculated in step f);

h) drawing a random sample from the posterior probability distribution to provide a simulated value for the primary variable;

i) entering the simulated value of said primary variable into said array at said selected station as an additional control pixel;

j) repeating steps e) through i) until values for the primary variable are simulated for all pixels of the array to define the spatial distribution of the simulated primary variable over the array of pixels;

k) with the aid of a computer graphics program, displaying a map of the spatial distribution of the simulated primary variable over the area of interest.

10. The method as defined by claim 9, comprising:

repeating steps e) through j) n times to provide n equally likely models of the spatial distribution of the simulated primary variable over the area of interest;

averaging the n equally likely models; and

executing step k) using the averaged model.

13. An article of manufacture, comprising:

a tangible machine-readable medium that embodies a set of instructions for programming a computer to execute the process steps defined by claim 10.