1. OBTAINING LOCAL PROPORTIONS FROM INVERTED SEISMIC DATA TOWARD PATTERN-BASED DOWNSCALING OF SEISMIC DATA
Lisa Stright and Alexandre Boucher School of Earth Sciences STANFORD UNIVERSITY
Main goals to get across:
Subseismic lithofacoeis prediction from multiple-attributes with an explicit model of scale and stationarity
Use multi-scale calibration and forward modeling to understand which lithofacies at which scale we consistently predict. Use to better understand spatial relationships of faices for modeling and interpretation.

2. Multiple-point geostatistics - SNESIM
A = Categorical Variable
B = Training image
C = Seismic Probability
P(A = channel | B = TI ) = 4/5 = 80%
P(A = non-channel | B = TI ) = 1/5 = 20%
Journel, 1992;
Guardiano and Srivastava, 1992;
Strebelle, 2000, 2002

3. Multiple-point geostatistics with soft data
A = Categorical Variable
B = Training image
C = Seismic Probability
P( A = channel | B = TI ) = 4/5 = 80%
P( A = non-channel | B = TI ) = 1/5 = 20%
P( A = channel | C = Seismic ) = 70%
Seismic Attribute
Probability 1 1
P( A | B, C )
- Combine with Tau Model
- Use dual training images

4. Scaling and probabilities?
Seismic Attribute
47%
20%
PSand
#1 #2 #3 1
Probability
Describe how the calibration can be used to predict probabilities away from well locationsS
Seismic Attribute
Data
Calibration
Realization(s)

5. after Campion et al., 2005; Sprague et al., 2002, 2006
Assumptions – Scale???
Probabilities and Facies can be scaled to the model grid
Seismic informs a homogeneous package
Homogeneous package can be represented by “most of” facies upscaling in wells
Probabilities account for inexact relationship between wells and seismic attribute(s)
Well 10 20 30 40 50 60 70 80 90
100
110
120
130
140
150
160
170
180
190
10’s of meters
(10’s)meters
Seismic
after Campion et al., 2005; Sprague et al., 2002, 2006
~ 100 m
Meters to 10’s of meters
1 m
Model scale ?

6. Proposed approach or methodology
Assumptions challenged when:
System is heterolithic (more than two categories)
Heterogeneities are smaller than seismic resolution (always?)
Multiple seismic attributes lumped into probabilities
Proposed Solution:
Create a multi-scale, multi-attribute well to seismic calibration
Use calibration to obtain local facies proportions at each seismic voxel location
Advantages of proposed approach
Can use any number of seismic attributes
Not dependent upon forward modeling (but can leverage forward modeling)
Uncertainty in tie between data types
Considers underlying cause of fine scale heterogeneity on coarse scale measurement response
Powerful when combined with knowledge of data (rock physics response, depositional setting and patterns)

7. Local Proportions from seismic attributes?
Describe how the calibration can be used to predict probabilities away from well locationsS
Seismic Attribute #1
Directly from calibration
From forward modeling
Data
Calibration
Realization(s)

8. Validation: Upper Cretaceous Cerro Toro Formation, Magallanes Basin
WL1
Mud Matrix Supported - Top of Slurry
Sandy Matrix Supported Top of Conglomerate
Debris Flow
WL2
Clast Supported Conglomerate
Clast Supported Conglomerate - Base of Slurry
WL3
Thin Beds - Sandstone/Mudstone/Conglomerate
Fine Grained Sandstone
Medium Grained Sandstone
Coarse Grained Sandstone
WL4
Thin Beds - Sandstone/Mudstone
WL5
Mud

9. Wildcat Lithofacies Channel fill Out-of-channel
Clast supported conglomerate
Conglomeratic mudstone
Thick bedded sandstone
Out-of-channel
Interbedded sandstone & mudstone
Mudstone with thin sand interbeds
LOOK AT DOM ON THIS PIC

10. Rock Properties: Late Oligocene Puchkirchen Formation, Molasse Basin, Austria
Bierbaum 1
17km
10km
AI (g/cm3m/s)
5000
13000

11. Multi-scale, multi-attribute calibration
1.4
1.5
1.6
1.7
1.8
1.9 2
2.1
2.2 4 6 8 10 12
1.7
1.8
1.9 2
2.1
2.2 6 8 10 12
1.4
1.5
1.6 4
Vp / Vs
Acoustic Impedance (g/cm3 m/s)

12. Create synthetic properties: Markov Chains
1.4
1.5
1.6
1.7
1.8
1.9 2
2.1
2.2 4 6 8 10 12
Vp / Vs
Acoustic Impedance (g/cm3 m/s)
Synthetics
1.7
1.8
1.9 2
2.1
2.2 6 8 10 12
1.4
1.5
1.6 4
0.8
0.043
0.106
0.014
0.029
0.002
0.006
0.083
0.781
0.005
0.01
0.115
0.018
0.952
0.003
0.011
0.012
0.956
0.978
0.001
0.022
0.004
0.989
0.017
0.983
0.008
0.984
0.962

13. Forward and Inverse Modeling
WL1
Mud Matrix Supported - Top of Slurry
Sandy Matrix Supported Top of Conglomerate
Debris Flow
WL2
Clast Supported Conglomerate
Clast Supported Conglomerate - Base of Slurry
WL3
Thin Beds - Sandstone/Mudstone/Conglomerate
Fine Grained Sandstone
Medium Grained Sandstone
Coarse Grained Sandstone
WL4
Thin Beds - Sandstone/Mudstone
WL5
Mud
50 Hz
15 Hz
25 Hz

14. Realizations
ThinBeds(s)
Sandstones(s)
Conglomerate(s)

15. Validation using field data
240
260
280
300
320
340
360
380
Proportion
400
420 1
Realization # 50

16. Outcrop results: Local Proportions
Prediction “good” when mean bed thickness
is at least 1/10 of seismic resolution
WL1
Mud Matrix Supported - Top of Slurry
Sandy Matrix Supported Top of Conglomerate
Debris Flow
WL2
Clast Supported Conglomerate
Clast Supported Conglomerate - Base of Slurry
WL3
Thin Beds - Sandstone/Mudstone/Conglomerate
Fine Grained Sandstone
Medium Grained Sandstone
Coarse Grained Sandstone
WL4
Thin Beds - Sandstone/Mudstone
WL5
Mud

17. Subsurface Application: Single Well
13000
6000

18. Subsurface application: log validation
Proportion
Realization # Is
Vp/Vs Ip

19. Subsurface Application: Single Well
13000
6000 1

20. Stratigraphic Layer 3 Prop( Conglomerate | Ip, Is, Vp/Vs )
Prop( ThinBeds | Ip, Is, Vp/Vs )
Prop( Sand | Ip, Is, Vp/Vs )
Prop( Mud/Disturbed | Ip, Is, Vp/Vs )

21. Compiling patterns from each layer

22. Summary and Conclusions
Multi-scale, multi-attribute calibration
Extract more information from well to seismic calibration to define inhomogeneous seismic “packages”
Explicitly handling scale differences in data to get full information content of each data source
Aid in calibrating inexact relationship between wells and seismic
Facies from wells/core
Multiple attributes from seismic
Gaps of unsampled events filled with forward modeling
Proportions and stacking patterns (vertical and lateral) need to be considered together
Underlying “patterns” linked to better search uncertainty space

23. Future Work Methodology Validation with Outcrop Models
What is the effect of seismic resolution and/or noise on the predictions?
What controls when a proportion set is prediction correctly?
Number of facies?
Bed thicknesses?
Stacking patterns?
Surrounding facies?
Calibration and Realizations
More intelligent selection of proportions based on spatial relationship with adjacent cells
Leverage the tie between the proportion and the underlying “pattern”
Determine which proportions are consistently predicted with multiple realizations and “freeze”
Analyze to better understand seismic “packages”
Remaining components defined by the model (Training Image)
Training Image generation and modeling

24. Acknowledgements
Industry Sponsor: Richard Derksen and Ralph Hinsch (RAG)
SPODDS Students: Dominic Armitage, Julie Fosdick, Anne Bernhardt, Zane Jobe,
Chris Mitchell, Katie Maier,
Abby Temeng,Jon Rotzien,
Larisa Masalimova
Advising Committee:
Stephen Graham, Andre Journel, Gary Mavko, Don Lowe Alexandre Boucher

25. References
Arpat, G. B., and Caers, J., 2007, Conditional simulation with patterns, Mathematical Geology, v. 39, no. 2, p  
Chugunova, T. L., and Hu, L. Y., 2008, Multiple-Point Simulations Constrained by Continuous Auxiliary Data, Mathematical Geosciences, v. 40, no. 2, p  
González, E. F., Mukerji, T., and Mavko, G., 2008, Seismic inversion combining rock physics and multiple-point geostatistics, Geophysics, v. 73, p. R11.  
Krishnan, S., 2008, The Tau Model for Data Redundancy and Information Combination in Earth Sciences: Theory and Application, Mathematical Geosciences, v. 40, no. 6, p  
Liu, Y., and Journel, A. G., 2008, A package for geostatistical integration of coarse and fine scale data, Computers and Geosciences.
Strebelle, S., 2002, Conditional simulation of complex geological structures using multiple-point statistics, Mathematical Geology, v. 34, no. 1, p  
Stright, L., 2006, Modeling, Upscaling, and History Matching Thin, Irregularly-Shaped Flow Barriers: A Comprehensive Approach for Predicting Reservoir Connectivity, SPE , in Proceedings SPE Annual Technical Conference and Exhibition, ATCE.
Stright, L., Stewart, J., Farrell, M., and Campion, K. M., 2008, Geologic and Seismic Modeling of a West African Deep-Water Reservoir Analog (Black’s Beach, La Jolla, Ca.) (abs.), in Proceedings American Association of Petroleum Geologists Annual Convention, Abstracts with Programs, San Antonio, Texas.
Zhang, T., Switzer, P., and Journel, A., 2006, Filter-based classification of training image patterns for spatial simulation, Mathematical Geology, v. 38, no. 1, p