Master geostatistical workflow card

This is the closing reference card for the Geostatistics textbook. Walk through every reservoir-characterisation project, clastic, carbonate, fluvial, tight-gas, heavy-oil EOR, CO₂ storage, by applying the decision graph below. Each row is a workflow stage; the right column is the textbook section.

Stage 1, Frame the decision question

• What is the DECISION RIDING ON THIS WORK? OOIP for development go/no-go? Well-spacing for drainage planning? CO₂ storage approval?
• What is the QUANTITY OF INTEREST? Mean (OOIP), connectivity (drainage), extreme (P-failure), uncertainty range (P90/P10)?
• What are the THRESHOLDS that distinguish acceptable from unacceptable outcomes? Document these BEFORE the analysis.

Stage 2, Assemble and prepare the data

• Well-log data: porosity, permeability, facies indicator, water saturation.
• Seismic-derived attributes: impedance, AVO products, time-lapse signatures (§4.5 cosimulation context).
• Declustering (§§2.1-2.3): for irregular well spacing, compute declustering weights before univariate statistics.
• Normal-score transform (§1.4): on the declustered data, NOT the raw data.

Stage 3, Build the spatial-correlation model

• Experimental variogram (§3): compute on the declustered, NS-transformed data; multiple directions for anisotropy detection.
• Variogram model fit (§4): spherical / exponential / Gaussian; permit nested structures; flag the nugget effect.
• Sensitivity: rerun downstream simulations with alternative variogram parameters to bound the variogram-model uncertainty.

Stage 4, Choose the estimation method

• Continuous Gaussian property + linear interpolation goal: ORDINARY KRIGING (§5.2).
• Need uncertainty per cell + many realizations: SEQUENTIAL GAUSSIAN SIMULATION (§§7.1-7.4).
• Categorical / facies modeling: indicator simulation (§§8.2-8.5).
• Channels, complex geometry, multipoint patterns: MPS / SNESIM (§§9.2-9.3).
• Modern generative ML (§9.4): only when validated against MPS and conditioning-data fidelity proven.
• Method choice driven by THREE axes (§9.5): variable type, spatial structure, data context.

Stage 5, Run, validate, and quantify uncertainty

• Number of realizations: 20+ for OOIP-type integrated metrics; 100+ for extreme-event probabilities (CO₂ P-failure, fracture-network connectivity).
• Validate each realization against the variogram, against well data exact-honoring, and against geological plausibility.
• Cross-validate via leave-one-out (§6.1-6.3): kriging-variance vs observed mean-squared error.
• Compute the decision-relevant quantity across realizations: P10/P50/P90 of OOIP, drainage area, P-failure.

Stage 6, Integrate auxiliary data

• Cosimulation with seismic attributes (§5.3 KED, §6 cokriging): dense secondary data constrains realizations between wells.
• For 4D / EOR (§10.5): time-lapse seismic informs sweep history; use Bayesian/ensemble methods to update realizations against observed 4D signatures.
• Geomechanical / fracture data (§10.4): integrate as secondary input to facies indicator or as separate fracture-density realization.

Stage 7, Communicate

• Report P10 / P50 / P90 of the decision-relevant quantity.
• State the VARIOGRAM MODEL explicitly: range, sill, nugget, anisotropy. Show alternative models that bracket the range.
• State the CONDITIONING-DATA fidelity: did realizations exactly honor wells? If MPS, was conditioning successful or approximate?
• For high-stakes claims (CO₂ approval, billion-dollar development): show the sensitivity of conclusions to variogram parameters AND to the choice of estimation method.
• NEVER report a single deterministic kriging map as if it captures uncertainty, kriging is the conditional mean only.

The two big traps to avoid

• Single-realization decisions: presenting kriging as "the answer". Kriging is the MEAN; the SPREAD across SGS realizations is the uncertainty. Report both, always.
• Wrong-method choice (§9.5): SGS on a fluvial reservoir misses channel connectivity; MPS on a homogeneous-shelf reservoir overspecifies. Match the method to the geology.

Closing thought

Reservoir characterisation is a CRAFT. The math (kriging equations, variogram models, MPS templates) is necessary but not sufficient. You must integrate it with geology, geophysics, reservoir engineering, and a clear understanding of what decisions the work will inform. The best geostatistical workflow is the one that gives the development team an honest distribution of outcomes, including the uncertainty they don't want to see.

You have completed the Geostatistics curriculum. Apply this material to real projects; teach it to colleagues; revisit it when new data acquisition technologies emerge (4D seismic, fibre optics, microseismic monitoring, AI-derived facies probabilities). The field will evolve, but the foundational ideas, variograms, kriging, SGS, MPS, uncertainty as ensemble, will remain.

Mark workflow card complete →

References, the close-of-textbook reading list

• Pyrcz, M.J., Deutsch, C.V. (2014). Geostatistical Reservoir Modeling, 2nd ed. Oxford. (The practitioner's comprehensive reference.)
• Deutsch, C.V., Journel, A.G. (1998). GSLIB: Geostatistical Software Library, 2nd ed. Oxford. (The canonical algorithms reference.)
• Chilès, J.-P., Delfiner, P. (2012). Geostatistics: Modeling Spatial Uncertainty, 2nd ed. Wiley. (The theoretical-rigorous reference.)
• Caers, J. (2011). Modeling Uncertainty in the Earth Sciences. Wiley. (Methodological framework, beautifully written.)
• Mariethoz, G., Caers, J. (2014). Multiple-point Geostatistics. Wiley-Blackwell. (MPS methods.)
• Goovaerts, P. (1997). Geostatistics for Natural Resources Evaluation. Oxford. (Classical reference, still indispensable.)
• Cressie, N. (1993). Statistics for Spatial Data. Wiley. (The statistical-theoretical reference for spatial statistics.)
• Wackernagel, H. (2003). Multivariate Geostatistics, 3rd ed. Springer. (Cokriging and multivariate methods.)
• Isaaks, E.H., Srivastava, R.M. (1989). An Introduction to Applied Geostatistics. Oxford. (Excellent introductory reference.)
• Rossi, M.E., Deutsch, C.V. (2014). Mineral Resource Estimation. Springer. (Mining-specific geostatistics applications.)