22.3.3.3 Data Analysis
The next important step in property modelling is data
analysis. The blocked well data (and other data if
relevant) are analysed using geostatistical techniques
in order to be able to set up the modelling process
optimally. The analysis includes:
• Histogram plotting, which shows the data distribution of a given property (Fig. 22.11a).
• Scatter plots of the properties versus spatial
dimensions. This will show if the property has
trends that are important to acknowledge in the
modelling. One trend typically found is a porosity
decrease with depth (Fig. 22.11b).
• Crossplotting of various properties (multivariate
analysis). Commonly, a correlation between porosity and the logarithm of permeability is found
(Fig. 22.11c).
• Spatial correlations are found by variogram analysis. This is a common technique in geostatistics,
analysing how the property will vary in a spatial
context (Fig. 22.11d).
The data analysis provides the fundamental input to
both facies modelling methods and modelling of continuous properties.
22.3.3.4 Facies Modelling
The word “facies” has several different definitions in
geology; for reservoir modelling, a facies is an architectural element that has a certain geometry and a
properly defined distribution of its internal continuous
properties. As an example, a fluvial reservoir zone
may comprise channel facies with high porosities
and a quite specific elongated shape, while the
overbank facies has poor reservoir properties with no
particular shape (only filling the space between channel bodies). The boundary between each facies may be
quite sharp, see example in Fig. 22.9. It is important to
understand that facies modelling is an optional step; is
not required in order to get a model that can be used
for volumetrics and flow. However, a facies model has
the advantage of introducing geologically reasonable
shapes into the geomodel, thus honouring the conceptual geological understanding. Most reservoir models
utilise facies modelling in some sense in order to
improve the petrophysical architecture, although complex facies models are demanding to construct. The
important choice of whether to do facies modelling or
not should be based on the modelling objectives, the
conceptual model, the need for accuracy and the available time and resources.
Whether facies is modelled in a simple fashion (e.g.
just two or three facies) or more complex facies patterns
depends on the conceptual model and model purpose.
Models that should be input to dynamic simulation
usually require facies models as the facies dictates
flow units and heterogeneity which can have significant
impact on fluid movements in the subsurface.
From a model technical view there are several
available techniques for making a facies model. The
main techniques, with a short description of pros and
cons, are briefly mentioned here:
Pixel-Based Facies Modelling In contrast to objectbased methods, the pixel-based methods use the 3D
grid directly in the modelling. A number of different
pixel based methods are available.
• Truncated Gaussian fields (Deutsch 2002, Chap. 6).
This methodology is useful for modelling large
scale facies objects that appear in a specific order,
like the boundaries between the lower, middle and
upper part of a shoreface, or the boundary between
shoreface and delta plain (Fig. 22.12a). Truncated
Gaussian fields are usually unsuitable for flow units
which have distinct shapes.
• Sequential indicator simulations (Deutsch 2002,
Chap. 6). This methodology can output a large
number of random patterns and has the advantage
of quite fast modelling when conditioning to a large
number of wells. However, the output from indicator methods may have difficulties in matching the
conceptual geological model.
• Multi-point statistics (Deutsch 2002, Chap. 11).
This technique utilises training images, providing
an intuitive way for the geologist to describe the
geology. This methodology has the advantage of
quite fast modelling when conditioning to a large
number of wells, and may in some cases replace the
need for using object-based methods.
Object-Based Facies Modelling This is a class of
facies modelling techniques that utilises a statistical
marked-point process to create facies objects based on
geometrical input (Deutsch 2002, Chap. 7). Object
22 Introduction to Reservoir Modelling
571
The next important step in property modelling is data
analysis. The blocked well data (and other data if
relevant) are analysed using geostatistical techniques
in order to be able to set up the modelling process
optimally. The analysis includes:
• Histogram plotting, which shows the data distribution of a given property (Fig. 22.11a).
• Scatter plots of the properties versus spatial
dimensions. This will show if the property has
trends that are important to acknowledge in the
modelling. One trend typically found is a porosity
decrease with depth (Fig. 22.11b).
• Crossplotting of various properties (multivariate
analysis). Commonly, a correlation between porosity and the logarithm of permeability is found
(Fig. 22.11c).
• Spatial correlations are found by variogram analysis. This is a common technique in geostatistics,
analysing how the property will vary in a spatial
context (Fig. 22.11d).
The data analysis provides the fundamental input to
both facies modelling methods and modelling of continuous properties.
22.3.3.4 Facies Modelling
The word “facies” has several different definitions in
geology; for reservoir modelling, a facies is an architectural element that has a certain geometry and a
properly defined distribution of its internal continuous
properties. As an example, a fluvial reservoir zone
may comprise channel facies with high porosities
and a quite specific elongated shape, while the
overbank facies has poor reservoir properties with no
particular shape (only filling the space between channel bodies). The boundary between each facies may be
quite sharp, see example in Fig. 22.9. It is important to
understand that facies modelling is an optional step; is
not required in order to get a model that can be used
for volumetrics and flow. However, a facies model has
the advantage of introducing geologically reasonable
shapes into the geomodel, thus honouring the conceptual geological understanding. Most reservoir models
utilise facies modelling in some sense in order to
improve the petrophysical architecture, although complex facies models are demanding to construct. The
important choice of whether to do facies modelling or
not should be based on the modelling objectives, the
conceptual model, the need for accuracy and the available time and resources.
Whether facies is modelled in a simple fashion (e.g.
just two or three facies) or more complex facies patterns
depends on the conceptual model and model purpose.
Models that should be input to dynamic simulation
usually require facies models as the facies dictates
flow units and heterogeneity which can have significant
impact on fluid movements in the subsurface.
From a model technical view there are several
available techniques for making a facies model. The
main techniques, with a short description of pros and
cons, are briefly mentioned here:
Pixel-Based Facies Modelling In contrast to objectbased methods, the pixel-based methods use the 3D
grid directly in the modelling. A number of different
pixel based methods are available.
• Truncated Gaussian fields (Deutsch 2002, Chap. 6).
This methodology is useful for modelling large
scale facies objects that appear in a specific order,
like the boundaries between the lower, middle and
upper part of a shoreface, or the boundary between
shoreface and delta plain (Fig. 22.12a). Truncated
Gaussian fields are usually unsuitable for flow units
which have distinct shapes.
• Sequential indicator simulations (Deutsch 2002,
Chap. 6). This methodology can output a large
number of random patterns and has the advantage
of quite fast modelling when conditioning to a large
number of wells. However, the output from indicator methods may have difficulties in matching the
conceptual geological model.
• Multi-point statistics (Deutsch 2002, Chap. 11).
This technique utilises training images, providing
an intuitive way for the geologist to describe the
geology. This methodology has the advantage of
quite fast modelling when conditioning to a large
number of wells, and may in some cases replace the
need for using object-based methods.
Object-Based Facies Modelling This is a class of
facies modelling techniques that utilises a statistical
marked-point process to create facies objects based on
geometrical input (Deutsch 2002, Chap. 7). Object
22 Introduction to Reservoir Modelling
571