• Handling of variance from multiple-scale
datasets is frequently incorrect;
• The tool-set for upscaling is still incomplete
and far from integrated; for example multiphase
flow, gridding and fault seal are generally
treated in separate software packages and
require a degree of manual data-file conversion.
Software tool developments will undoubtedly
steadily resolve these challenges, but what ulti-
mately is the goal? We suggest the overall target
of reservoir modelling is multi-scale (pore-to-field)
modelling and data integration. The level of detail
involved depends very much on the task at hand.
Some problems are essentially pore-scale - e.g.
will a different fluid displacement mechanism
such as CO 2 injection make a difference to ulti-
mate oil recovery? Other problems are essentially
large-scale - e.g. does this gas field have sufficient
volumes to justify a billion dollar investment?
Nevertheless, executing either of these projects in
detail will require a multi-scale analysis.
numerical upscaling method to determine the
upscaled flow property. The upscaled flow prop-
erty is then used as input in the next scale up. A
realistic illustration of this workflow for the pore
to lithofacies scale is shown in Figure 4.32 . Here
we assume we can define two different pore/rock
types (e.g. coarse well sorted sand and fine-
Flow functions for each rock type are defined
either from Special Core Analysis (SCAL) or
from pore network modelling, or preferably
both. Secondly, we assume we have a selection
of different facies models: e.g. trough cross bed-
ded sandstone (as in Fig. 4.32 ). These models
should correspond to the selection of modelling
elements described in Chapter 2.4. For each
lithofacies element, pore-scale properties are
assigned to each lamina (or bed). Upscaling is
performed to calculate the lithofacies-scale flow
properties (absolute and relative permeabilities
for each flow direction). These flow properties
are then assigned to each cell in the geomodel,
with further upscaling to the reservoir simulator,
For most cases, to make this explicit pore-to-
field upscaling computationally feasible, we use
steady-state approximations to multi-phase flow
(e.g. capillary limit and viscous limit methods).
These steady-state approximations have been
reviewed and discussed by for example Ekran
and Aasen ( 2000 ) and Pickup and Stephen
( 2000 ). Published examples of the pore to
lithofacies to full-field multi-scale workflow
include Pickup et al. ( 2000 ), Theting et al.
( 2005 ) and Rustad et al. ( 2008 ).
4.4.2 Pore-to-Field Workflow
There are many ways for defining a series of
explicit steps from the pore scale to the full-
field scale. The following summarises a typical
geologically-based workflow within a multi-
scale design framework. We define four domi-
nant length scales:
• Pore scale:
• Lithofacies scale: cm-m scale
10 m-10 km scale
• Reservoir simulator scale (typically some
coarsening up of the full-field reservoir
geomodel): 100 m-10 km scale.
These scales are based both on the nature of
rock heterogeneity and the principles for
establishing macroscopic flow properties. These
four scales give three transitions:
1. Pore to lithofacies
2. Lithofacies to geomodel
3. Geomodel to reservoir simulator.
At each scale we define flow properties for
each cell (or pore) in the model and then use a
4.4.3 Essentials of Multi-scale
We conclude this chapter with a check-list of
essential questions that need to be asked for any
reservoir flow-modelling problem.
1. Have you identified the main reservoir
elements that impact on flow?
• Hint: Use
the HFU concept
petrophysically distinct units