form, such as 0.25). The properties are not restricted to
traditional petrophysical properties; e.g. they may
include other physical properties such as geophysical
data (acoustic impedance, P wave velocity, etc.).
In order to achieve realistic results using a property
model, a mix of categorical and continuous properties
are used. Best known is the “two-stage” approach
applied for petrophysical properties. First, the main
flow units are modelled using a categorical property;
so-called facies modelling (Sect. 22.3.3.4). Next, each
facies is populated independently with continuous
petrophysical properties (Sect. 22.3.3.5). The principles
of going from a facies model to continuous properties
are illustrated in Fig. 22.9.
The next question is how shall the property model
be populated in practice? What input data are available? Generally speaking, we have the following data
available:
• Well data, in the form of a zone log, facies log, porosity, saturation and permeability logs. These are usually
treated as hard data, meaning that well observations
should be conditioned properly in a model.
• Seismic data, in the form of depth-converted
interpretations and attribute cubes. These data
have usually poorer resolution than the cells we
model. Seismic data are regarded as less certain
than well data, and the use of seismic data will
vary a lot from field to field, depending on seismic
resolution and availability. In general, seismic data
should be treated as semi-quantitative.
• The conceptual model (see Sect. 22.2.3) may be
regarded as “soft” data, meaning that is cannot be
conditioned directly; rather the model is “tuned” to
acknowledge the main features in the concept
description.
22.3.3.1 The Role of Geostatistics
Before we proceed into actual property modelling
methods, it is important to emphasise the role of
geostatistics in reservoir modelling. Both the analysis
and modelling of facies and petrophysical properties
are heavily driven by geostatistical principles and
methods, and the use of geostatistical methods is also
increasing within structural framework modelling.
The field of statistics in general is concerned with
quantitative methods for collecting, summarising and
analysing data, as well as drawing conclusions and
making reasonable decisions on the basis of
such analysis, and is an extensive science.
Geostatistics is distinct from statistics in three main
aspects: (1) focus on the geological origin of the data,
(2) explicit modelling and treatment of spatial
correlation between data, and (3) treatment of data at
different volume scales and levels of precision
(Deutsch 2002).
The geomodeller will need a certain level of skills
within geostatistics as the modelling software draws
heavily upon the geo-statistical methods and terminology. A sufficient skill level is usually a good understanding of geostatistical principles and terms; the
detailed mathematics behind it is usually not required.
However, it is common that oil companies employ a
number of statistical experts who are available as
support when doing reservoir modelling in general,
and in particular the uncertainty analysis associated
with it (Sect. 22.5).
22.3.3.2 Blocking of Well Data
Before starting the actual property modelling, the
available well data need to be sampled into the scale
of the geological modelling grid. This process is most
commonly referred to as blocking of wells/well data
but could also be called upscaling of well data. This
involve sampling and averaging of petrophysical log
data into the scale of the geological grid.
Blocking of well data is not a trivial task. The well
logs (see Chap. 16) are a result of an averaging due to
the logging tools, particular measurement technique,
and the sampling increment is usually approximately
15 cm for conventional logs. The actual resolution is
usually less, typically from 30 cm to several metres,
depending on the tool and measurement conditions.
This means that small-scale laminae (from mm scale
to tens of cm scale) commonly seen in cores are not
captured by the logs, and shoulder bed effects (the
smearing or averaging when going from one lithology
to another) can be significant. When these logs are
blocked, a further averaging is done. Horizontal
wells have additional challenges. An example of well
blocking is shown in Fig. 22.10. Loss of detail from
the original logs may be compensated by using a finer
grid; however using too many grid cells will be a
disadvantage as it will increase the needed computation power later in the modelling process.
22 Introduction to Reservoir Modelling
569
traditional petrophysical properties; e.g. they may
include other physical properties such as geophysical
data (acoustic impedance, P wave velocity, etc.).
In order to achieve realistic results using a property
model, a mix of categorical and continuous properties
are used. Best known is the “two-stage” approach
applied for petrophysical properties. First, the main
flow units are modelled using a categorical property;
so-called facies modelling (Sect. 22.3.3.4). Next, each
facies is populated independently with continuous
petrophysical properties (Sect. 22.3.3.5). The principles
of going from a facies model to continuous properties
are illustrated in Fig. 22.9.
The next question is how shall the property model
be populated in practice? What input data are available? Generally speaking, we have the following data
available:
• Well data, in the form of a zone log, facies log, porosity, saturation and permeability logs. These are usually
treated as hard data, meaning that well observations
should be conditioned properly in a model.
• Seismic data, in the form of depth-converted
interpretations and attribute cubes. These data
have usually poorer resolution than the cells we
model. Seismic data are regarded as less certain
than well data, and the use of seismic data will
vary a lot from field to field, depending on seismic
resolution and availability. In general, seismic data
should be treated as semi-quantitative.
• The conceptual model (see Sect. 22.2.3) may be
regarded as “soft” data, meaning that is cannot be
conditioned directly; rather the model is “tuned” to
acknowledge the main features in the concept
description.
22.3.3.1 The Role of Geostatistics
Before we proceed into actual property modelling
methods, it is important to emphasise the role of
geostatistics in reservoir modelling. Both the analysis
and modelling of facies and petrophysical properties
are heavily driven by geostatistical principles and
methods, and the use of geostatistical methods is also
increasing within structural framework modelling.
The field of statistics in general is concerned with
quantitative methods for collecting, summarising and
analysing data, as well as drawing conclusions and
making reasonable decisions on the basis of
such analysis, and is an extensive science.
Geostatistics is distinct from statistics in three main
aspects: (1) focus on the geological origin of the data,
(2) explicit modelling and treatment of spatial
correlation between data, and (3) treatment of data at
different volume scales and levels of precision
(Deutsch 2002).
The geomodeller will need a certain level of skills
within geostatistics as the modelling software draws
heavily upon the geo-statistical methods and terminology. A sufficient skill level is usually a good understanding of geostatistical principles and terms; the
detailed mathematics behind it is usually not required.
However, it is common that oil companies employ a
number of statistical experts who are available as
support when doing reservoir modelling in general,
and in particular the uncertainty analysis associated
with it (Sect. 22.5).
22.3.3.2 Blocking of Well Data
Before starting the actual property modelling, the
available well data need to be sampled into the scale
of the geological modelling grid. This process is most
commonly referred to as blocking of wells/well data
but could also be called upscaling of well data. This
involve sampling and averaging of petrophysical log
data into the scale of the geological grid.
Blocking of well data is not a trivial task. The well
logs (see Chap. 16) are a result of an averaging due to
the logging tools, particular measurement technique,
and the sampling increment is usually approximately
15 cm for conventional logs. The actual resolution is
usually less, typically from 30 cm to several metres,
depending on the tool and measurement conditions.
This means that small-scale laminae (from mm scale
to tens of cm scale) commonly seen in cores are not
captured by the logs, and shoulder bed effects (the
smearing or averaging when going from one lithology
to another) can be significant. When these logs are
blocked, a further averaging is done. Horizontal
wells have additional challenges. An example of well
blocking is shown in Fig. 22.10. Loss of detail from
the original logs may be compensated by using a finer
grid; however using too many grid cells will be a
disadvantage as it will increase the needed computation power later in the modelling process.
22 Introduction to Reservoir Modelling
569