Procedures for Lithology Characterization and Probabilistic Upscaling (Curve "Blocking") Using Petrophysical and Core Data

Eslinger, Eric¹ (1) Eric Geoscience, Inc. and The College of Saint Rose, Glenmont, NY

A new "expert" system, ModeAssign, automatically assigns electrofacies to lithology groups (rock types). The electrofacies are developed using a probabilistic clustering procedure. Significant time and mental gymnastics previously were required to produce a reasonable interpretation of the lithology for each electrofacies. With ModeAssign, each electrofacies is automatically assigned to one of five lithologies: sandstone, siltstone, claystone ("shale"), limestone, or dolostone. Minimum requirements are that the initial clustering analysis include bulk density, neutron, and gamma ray curves as variables. The clastic lithologies are defined by binning mean GR values for each mode using user-defined gamma ray cutoffs. Carbonate lithology assignments are driven by apparent grain densities derived from RHOB vs NPHI crossplots (for each electrofacies) extrapolated to apparent zero porosity. Warnings appear when the logic shows internal conflict with known petrophysical properties. The auto-assigned lithologies can be overridden at any time.

Building upon ModeAssign, a bed thickness filter has been designed that permits undesirable "thin" beds (also produced during probabilistic clustering) to be eliminated. The eliminated thin beds are added to adjacent, thicker beds using a probabilistic-based similarity index. This is essentially a one-dimensional up-scaling (or, "curve blocking") to any desirable minimum bed thickness. Mean values for each upscaled bed are computed for any curve in the well log las file. When multiple-well clustering runs are used, this procedure permits all wells to be upscaled to the same minimum bed thickness using the same rules. Upscaled beds can then be correlated and the results used for reservoir or geocell modeling.