Next Generation Geological Modeling for Hydrocarbon Reservoir Characterization - Advances in Data, Methods, Models and Their Applications in Geoscience - page 219

and orientation. The constraint for an unbiased estimator is satisfied by maintaining
1



.



2.2 Definition of lithofacies and mapping of lithotype proportions
High-resolution geomodeling of lithofacies and reservoir properties begins by creating a
properly sealed structural framework and then integrating available data from cores, well-
logs and seismic surveys. Ideally the lithology model is based on the interpretation of
deposition facies from core description. However, for siliciclastic environments, the
lithologies are typically electro-facies based on wireline logs calibrated to a few core
descriptions. Finally, we often used petro-facies for carbonate reservoir as the primary
depositional facies are destroyed by post depositional diagenetic processes. Once the
lithology is created, facies are populated with characteristic petrophysical properties.
Micro-logs can identify very thin impermeable shale layers, as thin as 1-2 ft. In order to
maintain vertical communication and small scale heterogeneity in the model, particularly in
assessing the sweep efficiency of EOR applications (Maučec and Chambers, 2009), e.g. CO 2
flooding (Culham et al. , 2010), it is critical for well blocking to preserve small-scale facies
heterogeneities in each interval, assign them correctly to common lithotypes and prevent
inadvertently eliminating the essential geological information. The integrity of the blocking
and preservation of small scale features is directly proportional to the vertical cells size. If
small scale, thin bedding, is critical to sweep efficiencies, then a small vertical cell is required
in the model, often resulting in a large multi-million cell geological model, which may be
used without upscaling in the dynamic simulator.
Modeling complex geologic environments ( e.g. fluvial, deltaic) require the ability to control
vertical relationships and lateral relationships between the facies. In stratigraphic modeling,
which is done on an interval-by-interval basis, the task is to identify the depositional
environment and primary depositional facies. Each depositional environment is controlled
by physical processes of sedimentation and erosion, which requires the creation of internal
bedding geometry, e.g. layers representative of the depositional system. These layers act as
lines of internal correlation that affect the gathering of statistical information in variogram
computation and the distribution of properties in subsequent modeling steps. Once the
layering styles are specified for each interval, the well data are re-sampled (coarsened or
blocked) at the scale of the layers and a single property value assigned to each layer along
the wellbore. For continuous properties, DecisionSpace  Desktop Earth Modeling uses
standard averaging methods to assign a value to the gravity center of each grid cell (layer)
along the wellbore, biased to the lithology code. For discrete properties, coded by integers
( e.g. facies) the most commonly occurring facies code is chosen.
Conventional modeling approaches use the average or global facies proportions per interval
of interest, which implicitly assumes that the facies proportions are unrealistically the same
everywhere throughout the interval and therefore applying a constant lithotype proportion
curve (LPC) to the entire interval is inaccurate. Traditional techniques to introduce a
geological trend in the data usually require laborious creation of pseudo-wells or
application of a generic trend map. The use of a generalized trend map implies that the
geological continuity throughout the interval would be fairly similar. In other words, the
interpretation of underlying statistics ( e.g. histograms, variograms) and characteristics such
as anisotropy and correlation length, would in mathematical terms, assume the condition of
stationarity (Caers, 2005). However, most reservoirs are non-stationary and the introduction