From Facies to Flow Units: Getting from Theory to Practice

Charles T. Feazel
ConocoPhillips, Houston, Texas

You’re focused on lithofacies, fracture patterns, or diagenetic histories. Your manager wants production forecasts and project economics. How do you get from here to there?

The first decision in building a reservoir model involves data availability: do you have sufficient data to be deterministic, or will you use sparse data or analogs and distribute reservoir properties stochastically? Then comes a second challenge: at what scale do you preserve reservoir heterogeneity through upscaling? Answers to these questions ultimately determine how closely the resulting models approximate subsurface geometries and flow characteristics.

The focus of this Hedberg Conference is interdisciplinary communication. Only by mutual understanding among geologists, geophysicists, petrophysicists, and reservoir engineers can we hope to capture the workflow synergies that make reservoir characterization truly useful in exploring for – and producing from – carbonate reservoirs. Limestones and dolostones exhibit properties sufficiently different from siliciclastic reservoirs that their characterization may require differing workflows. In particular, they may have:

• more variable stratal architecture, 
• increased vertical and lateral heterogeneity, 
• more complex pore networks and diagenetic histories, 
• different responses to structural deformation,
• all leading to unique porosity distributions and permeability anisotropies

Reservoir models have time-value: they can be constructed and applied to economic decisions very early in the life of an asset (pre-drill), or very late (production cessation and abandonment). They are valuable decision-making tools during primary, secondary, and tertiary recovery. They also help quantify the value of data at various points in the life of an asset. However, geomodel construction is not always straightforward. Workflows may become dictated by software, rather than by stratigraphic, structural, or geostatistical reasoning, and specific tasks like fault modeling can be time-intensive and non-intuitive.