1. 1 RPSEA Project – Facies probabilities from seismic data in Mamm Creek Field Reinaldo J Michelena Kevin Godbey Patricia E Rodrigues Mike Uland April 6, 2010

2. 2 Location of Mamm Creek field Mamm Creek

3. 3 Area of interest Reservoir simulation area Geological model area

4. 4 Stratigraphic column Coastal /Alluvial Plain Shoreface/Deltaic Marsh, Mire, Swamp, and Estuarine Marine Shelf/Ramp ~2,200 ft Geological model interval

5. 5 Intervals of interest for seismic analysis Price Coal UMV MiddleSS Rollins Cozette Mancos seismic tracedensity log Lower amplitudes (fluvial) Higher amplitudes (transitional and marine) beginning of coal interval Seismic amplitudes

6. 6 Reservoir characterization at Mamm Creek MOTIVATION Effective development of Mamm Creek field requires detailed understanding of all geological features that control gas accumulation and connectivity GOAL Build geological models that capture the main geological complexities of the field Show how seismic data can be used to help mapping sand distribution in Mamm Creek field this presentation

7. 7 Challenges from rock physics diagnostics sands Vp Vs RHO Vs RHO Vp Potential difficulties: 1)Sands and shales show significant overlap in log scale crossplots of elastic properties 2)Presence of thin coal layers may mask “real” seismic response from surroundings sands and shales. shales coals

8. 8 Data inventory Seismic: – 3D, PP pre-stack, NMO corrected gathers – 3D PS_fast and PS_slow stack volumes (not part of same survey as PP data) Well: – 102 density logs. Most wells with Gamma Ray, Neutron Porosity and Resistivity – 3 sonic logs – 2 dipole sonic logs (one of them crossdipole) – Formation tops from Bill Barrett Corporation – 8 cores – 12 FMI logs

9. 9 Summary of seismic workflow Perform petrophysics/rock physics diagnostics QC seismic data Precondition pre-stack gathers for AVO analysis Interpret PP data Interpret PS data (consistent with PP interpretation) Invert PP pre-stack data for Vp, Vs and density Invert PS stack data for pseudo shear impedance (Zps) Generate velocity model that honors all marker and horizon information in PP and PS time Perform time to depth conversion of seismic derived information

10. 10 Inverted seismic volumes in depth RHOVpVs Zps_fastZps_slow low high UMV MIddleSS Z

11. 11 Attribute crossplots Inverted seismic attributes (3 from PP and 2 from PS) can be cross plotted in many different ways After examining various combinations of these five attributes, we decided to focus only on 6 of such combinations and assess the contribution of each attribute: – Vp vs Vs – Vp vs RHO – Vs vs RHO – Zps _fast vs Zps _slow – Vp vs Vs vs RHO – Vp vs Vs vs RHO vs Zps _fast vs Zps _slow

12. 12 Seismic attributes at well locations (color = log scale facies) Vp Vs Vp RHO Vs RHO thick sandbackground Zps fast Zps slow Transitional and marine interval

13. 13 Log scale attributes (color = lithology) sands Zps fast Zps slow Vp Vs Vp RHO Vs RHO Transitional and marine interval

14. 14 Seismic attributes at well locations (color = log scale facies) Vp Vs Vp RHO Vs RHO thick sandbackground Zps fast Zps slow Transitional and marine interval

15. 15 Probabilities from seismic attribute crossplots Channel No channel

16. 16 Probabilities extracted at well 14C-20, fluvial Sand Ave 2 2 2 3 5 flags Vp, VsVp, RhoVs,RhoVp, Vs, RhoVp, Vs, Rho, 3C Gamma Lithology Ray 30 wells used for calibration UMV_MKR (4766’MD) TOP_GAS (5306’MD) MKR1 (6235MD’) 100’

17. 17 Probability using Vp, Vs, RHO, Zps_fast and Zps_slow 102 wells used for calibration 0.60 Thick sand probability Stratigraphic slice

18. 18 Real vs estimated sand thickness (Vp, Vs, RHO and 3C) Prob cutoff = 0.18 cc=0. 68 Fluvial interval Thickness from log data (ft) Thickness from seismic data (ft) 30 wells used to calibrate seismic

19. 19 Thick sand flags at well locations

20. 20 Seismic derived probabilities at well locations Probabilities from 5 attribute crossplot calibrated with 102 wells

21. 21 Thick sand flags and probabilities at well locations Probabilities from 5 attribute crossplot calibrated with 102 wells

22. 22 Correlation coefficient* per well (marine) Vp, Vs, RHO, PS, 30 calibration wells Vp, Vs, RHO, PS, 102 calibration wells Well name Correlation Vp, Vs, RHO, 30 calibration wells Vp, Vs, RHO, 102 calibration wells * Real average sand vs Estimated probability

23. 23 Sorted correlation coefficients (marine) Vp, Vs, RHO, 30 calibration wells Correlation Sort index

24. 24 Sorted correlation coefficients (marine) Vp, Vs, RHO, 30 calibration wells Vp, Vs, RHO, 102 calibration wells Correlation Sort index

25. 25 Sorted correlation coefficients (marine) Vp, Vs, RHO, PS, 30 calibration wellsVp, Vs, RHO, 30 calibration wells Vp, Vs, RHO, 102 calibration wells Correlation Sort index

26. 26 Sorted correlation coefficients (marine) Vp, Vs, RHO, 30 calibration wells Vp, Vs, RHO, 102 calibration wells Vp, Vs, RHO, PS, 30 calibration wells Vp, Vs, RHO, PS, 102 calibration wells Correlation Sort index

27. 27 Sorted correlation coefficients (fluvial) Vp, Vs, RHO, 30 calibration wells Vp, Vs, RHO, 102 calibration wells Vp, Vs, RHO, PS, 30 calibration wells Vp, Vs, RHO, PS, 102 calibration wells Correlation Sort index

28. 28 Log scale Vs _fast vs Vs _slow (well 41A-28) SandBackground Log scale Vs_fast (ft/s) Log scale Vs_slow (ft/s) Log scale Vs_fast (ft/s) Log scale Vs_slow (ft/s) Fluvial intervalMarine interval Fluvial shales are more azimuthally anisotropic than marine shales Fluvial sands are less azimuthally anisotropic than marine sands

29. 29 Conclusions Facies probability estimation algorithm from multidimensional crossplots of seismic attributes yields useful results even when elastic properties of sandy and background facies overlap completely When using PP data only, the best results are obtained when using Vp, Vs, and RHO simultaneously The best separation is achieved when using all five attributes (Vp, Vs, RHO, Zps_fast and Zps_slow) Sensitivity of PS data to azimuthal anisotropy helps to improve sand identification where sands are more anisotropic than the background

30. 30 Acknowledgments The authors acknowledge RPSEA (Research Partnership to Secure Energy for America) for financial support Thanks also to Bill Barrett Corporation for providing the data used for this study