Geodip K M Geoscience Ltd, geological consultants, measurement of geological dip, Aberdeen Scotland

It sounds simple, but modelling of porosity is perhaps the most important parameter to estimate during the modelling process. The best starting point is core plug data as this measurement refers to the pores that can be occupied by Helium (all pores sizes, but not bound water, -OH). The Helium porosity can be adjusted to reservoir conditions by allowing for overburden.

Are the core plugs representative?

Are they on depth with wireline logs?

The wireline logs measure porosity by a variety of techniques, but all are affected by rock density, bound water and the borehole conditions to some extent. An environmentally adjusted density - neutron combination log will provide a good measurement in a clean lithology. The petrophysicist will attempt to eliminate the bound water (“shale”) from the readings and calculate an effective porosity that can be calibrated to core plug data. This is sometimes known as effective porosity, but the definition varies somewhat from person to person. For purposes of calculating static volumes and hydrocarbons-in-place, the effective porosity is a good estimate (See Slideshow, Slide 7).

Are there any wells where the porosity logs look suspect?

A quick model with a few layers may identify wells where there is a shift that cannot be explained geologically

By what method was the porosity calculated and is it a consistent for all wells?

The common errors are inconsistent fluid or rock densities

Has the calculated porosity been adjusted to core plug data (where present)?

Thus may or may not be required

Are there any thin alternating sand and shale intervals where the logs are not fully responding to the sand beds?

Permeability is obtained primarily from core analysis data. In case there is no core, then analogue field data should be used. Routine core analysis data is measured as permeability to air at surface conditions. At some point in the workflow of the petrophysicist, geologist or engineer, this needs to be adjusted to liquid permeability at reservoir conditions.

Estimation of permeability in the uncored intervals is always subject to uncertainty. This is where identifying the facies can have an important influence on the final permeability distribution. The most useful method is the permeability- porosity cross plot of core plug data as this is not affected by log-core depth shifts. A porosity – permeability relationship can often be derived, but in some cases the scatter is so wide that it is meaningless. Other logs that can be used to estimate permeability are GR and density-neutron separation or NMR logs. Bear in mind that when using exponential relationships, a change in porosity from say 0.22 to 0.24 can result in a large change in the calculated permeability, so correct porosity estimate is important, with a check needed for extreme values.

When deriving permeability from porosity a statistical relationship is recommended as it is often more representative of the actual heterogeneity and should give better dynamic results. A method such as co-kriging can be used with a correlation coefficient of 0.7 for instance. This is different to the fixed relationships used traditionally by reservoir engineers.

Where are the high permeability intervals?

How far laterally are these expected to go?

Does well test data fit the calculated permeability (perm*height)?

Are high permeability intervals related to geological facies?

Net Reservoir

Net reservoir is that where a significant proportion of fluid is expected to flow under the reservoir conditions. That infers a permeability cut-off, but as permeability cannot be directly measured, other cut-offs are frequently used, such as porosity < 0.1 or Vshale > 0.4.

The wells can be subdivided into net and non-net intervals accordingly. A particular zone on a log may have a NTG of say 0.6. Unlike traditional mapping, it is not essential to map this NTG as a continuous property and often is causes confusion. It is better to model net and non-net facies (0/1) or use a cut-off after the modelling has been done, otherwise the benefits of the 3-D modelling are diluted.

There is another stage in the dynamic realm for the engineers to consider. Not all the net porosity is occupied by hydrocarbons that can move. That is where special core analysis parameters can be used to estimate pore size distribution and capillary pressure. From this parameters such as irreducible water and residual oil can be estimated that will help to calculate mobile oil under particular reservoir conditions.

The first property modelling step is to sample the well logs into the modelling grid at the well positions. If the layering is fine enough, then this is straightforward. It is best to make the non-net intervals in the well log curves indeterminate so that they are not included in the net cell values, i.e. for porosity and permeability. Avoid using special averaging methods, for a first pass at least, unless there is a particular strong reason. The key step is to compare the well log curves with the blocked well logs and check distribution histograms for net, porosity and permeability.

If thicker cells are required, often due to the number of cells that can be handled in the dynamic realm, then care must be taken that the log readings are not averaged out too much, resulting in lack of clear cut property layers and often leading to an underestimate of overall permeability in the reservoir.

Don’t forget, the reservoir engineering model behaves according to porosity and permeability, facies are not used. Extensive facies modelling is useless for dynamic work unless it reflects changes in permeability.

3-D Petrophysical Modelling Methods and Algorithms.

The method selected will depend on the number of wells and their spacing. It is common for the wells to target the best reservoir and they will not always represent the whole hydrocarbon volume being modelled. The geological model or facies model will influence how the properties are extrapolated away from well data points. Note that changing the correlation will change the interval being extrapolated.

It is best to start with a simple method such as moving average to see if any trends are apparent. Kridging is a good method for identifying trends across the whole field and works well for thin beds that are believed to cover a large area such as shorefaces.

Beware of unwanted abrupt changes in properties at the boundary of facies. This can occur if the facies are modelled independently. It is possible to use the facies as a guide rather than modelling the properties separately (See Slideshow, Slide 8).

Sequential Gaussian Simulation uses the statistical distribution of the well data and this can be used to generate lateral heterogeneity or indeed smoother patterns. A secondary property can be used to influence the result, such as a seismic attribute or a previously generated property.

The main task in petrophysical modelling is to generate areal variations in properties that attempt to reflect the geology and honour the well data. At the same time, it is usually required for the non-net facies, porosity and permeability to have similar trends, but with an element of heterogeneity. Vastly differing trends in the various properties will cause problems dynamically (e.g. high porosity, very low permeability).

It is useful to compare the statistical distribution of the data for original well logs, blocked well logs and the 3-D model. There would normally be a close match between the original logs and blocked logs for each zone. However the 3-D model is designed to be dependant on well spacing, facies modelling, depth trends and it is NOT normally required for the statistics to be the same as the well data. It is possible to make comparisons within a particular area where properties are expected to match, such as within the channel belts of a reservoir, ignoring the poorer quality interchannel areas.

Hydrocarbon Saturation

In most reservoirs the hydrocarbon saturation can be readily calculated from the well logs and a hydrocarbon-water contact can be established from exploration wells with formation pressure tests (RFTs). Extending these observations to calculate the saturations in the rest of the reservoir requires further steps. Initial wells will obviously show original saturations, but as soon as production has started, subsequent wells will have higher contacts and flushed intervals so cannot be used for OOIP calculations. The hydrocarbon saturation is usually at a maximum in the crest of the structure and decreases significantly near the original oil-water contact. The oil saturation is also higher in good quality reservoir than poor quality reservoir. Therefore it is common practice to use a function that relates water saturation to height above the free water level to reservoir quality. Each cell in the model can then be evaluated in 3-D and water saturation calculated. Methods that should be considered are:

In both these method a geometrical property of height above the FWL needs to be generated in the model.

While most methods rely on establishing a contact and this is important for static hydrocarbon volumes, it should also be recognised that the interval near the contact will not contribute very much to the reserves. It is not a good idea to artificially deepen the contact, based on very low oil saturations, unless there are special circumstances. Core fluorescence data may assist in defining the oil-water contact, but in some reservoirs there may be good fluorescence below the current free water level based on RFT measurements.

Are the core plugs representative?

Are they on depth with wireline logs?

Are there any wells where the porosity logs look suspect?

A quick model with a few layers may identify wells where there is a shift that cannot be explained geologically

By what method was the porosity calculated and is it a consistent for all wells?

The common errors are inconsistent fluid or rock densities

Has the calculated porosity been adjusted to core plug data (where present)?

Thus may or may not be required

Are there any thin alternating sand and shale intervals where the logs are not fully responding to the sand beds?

Where are the high permeability intervals?

How far laterally are these expected to go?

Does well test data fit the calculated permeability (permheight)?*

Are high permeability intervals related to geological facies?

Net Reservoir

Are the properties expected to be smooth are have a lot of variation between wells?

How much dependence is there between net, porosity and permeability?

Hydrocarbon Saturation




			Petrophysical Modelling Porosity It sounds simple, but modelling of porosity is perhaps the most important parameter to estimate during the modelling process. The best starting point is core plug data as this measurement refers to the pores that can be occupied by Helium (all pores sizes, but not bound water, -OH). The Helium porosity can be adjusted to reservoir conditions by allowing for overburden. Are the core plugs representative? Are they on depth with wireline logs? The wireline logs measure porosity by a variety of techniques, but all are affected by rock density, bound water and the borehole conditions to some extent. An environmentally adjusted density - neutron combination log will provide a good measurement in a clean lithology. The petrophysicist will attempt to eliminate the bound water (“shale”) from the readings and calculate an effective porosity that can be calibrated to core plug data. This is sometimes known as effective porosity, but the definition varies somewhat from person to person. For purposes of calculating static volumes and hydrocarbons-in-place, the effective porosity is a good estimate (See Slideshow, Slide 7). Are there any wells where the porosity logs look suspect? A quick model with a few layers may identify wells where there is a shift that cannot be explained geologically By what method was the porosity calculated and is it a consistent for all wells? The common errors are inconsistent fluid or rock densities Has the calculated porosity been adjusted to core plug data (where present)? Thus may or may not be required Are there any thin alternating sand and shale intervals where the logs are not fully responding to the sand beds? Core plugs may be biased towards the best sands in interbedded sequences. Permeability Permeability is obtained primarily from core analysis data. In case there is no core, then analogue field data should be used. Routine core analysis data is measured as permeability to air at surface conditions. At some point in the workflow of the petrophysicist, geologist or engineer, this needs to be adjusted to liquid permeability at reservoir conditions. Estimation of permeability in the uncored intervals is always subject to uncertainty. This is where identifying the facies can have an important influence on the final permeability distribution. The most useful method is the permeability- porosity cross plot of core plug data as this is not affected by log-core depth shifts. A porosity – permeability relationship can often be derived, but in some cases the scatter is so wide that it is meaningless. Other logs that can be used to estimate permeability are GR and density-neutron separation or NMR logs. Bear in mind that when using exponential relationships, a change in porosity from say 0.22 to 0.24 can result in a large change in the calculated permeability, so correct porosity estimate is important, with a check needed for extreme values. When deriving permeability from porosity a statistical relationship is recommended as it is often more representative of the actual heterogeneity and should give better dynamic results. A method such as co-kriging can be used with a correlation coefficient of 0.7 for instance. This is different to the fixed relationships used traditionally by reservoir engineers. Where are the high permeability intervals? How far laterally are these expected to go? Does well test data fit the calculated permeability (permheight)?* Are high permeability intervals related to geological facies? Net Reservoir Net reservoir is that where a significant proportion of fluid is expected to flow under the reservoir conditions. That infers a permeability cut-off, but as permeability cannot be directly measured, other cut-offs are frequently used, such as porosity < 0.1 or Vshale > 0.4. The wells can be subdivided into net and non-net intervals accordingly. A particular zone on a log may have a NTG of say 0.6. Unlike traditional mapping, it is not essential to map this NTG as a continuous property and often is causes confusion. It is better to model net and non-net facies (0/1) or use a cut-off after the modelling has been done, otherwise the benefits of the 3-D modelling are diluted. There is another stage in the dynamic realm for the engineers to consider. Not all the net porosity is occupied by hydrocarbons that can move. That is where special core analysis parameters can be used to estimate pore size distribution and capillary pressure. From this parameters such as irreducible water and residual oil can be estimated that will help to calculate mobile oil under particular reservoir conditions. Blocking of Well Logs into the Modelling Grid The first property modelling step is to sample the well logs into the modelling grid at the well positions. If the layering is fine enough, then this is straightforward. It is best to make the non-net intervals in the well log curves indeterminate so that they are not included in the net cell values, i.e. for porosity and permeability. Avoid using special averaging methods, for a first pass at least, unless there is a particular strong reason. The key step is to compare the well log curves with the blocked well logs and check distribution histograms for net, porosity and permeability. If thicker cells are required, often due to the number of cells that can be handled in the dynamic realm, then care must be taken that the log readings are not averaged out too much, resulting in lack of clear cut property layers and often leading to an underestimate of overall permeability in the reservoir. Don’t forget, the reservoir engineering model behaves according to porosity and permeability, facies are not used. Extensive facies modelling is useless for dynamic work unless it reflects changes in permeability. 3-D Petrophysical Modelling Methods and Algorithms. The method selected will depend on the number of wells and their spacing. It is common for the wells to target the best reservoir and they will not always represent the whole hydrocarbon volume being modelled. The geological model or facies model will influence how the properties are extrapolated away from well data points. Note that changing the correlation will change the interval being extrapolated. It is best to start with a simple method such as moving average to see if any trends are apparent. Kridging is a good method for identifying trends across the whole field and works well for thin beds that are believed to cover a large area such as shorefaces. Beware of unwanted abrupt changes in properties at the boundary of facies. This can occur if the facies are modelled independently. It is possible to use the facies as a guide rather than modelling the properties separately (See Slideshow, Slide 8). Sequential Gaussian Simulation uses the statistical distribution of the well data and this can be used to generate lateral heterogeneity or indeed smoother patterns. A secondary property can be used to influence the result, such as a seismic attribute or a previously generated property. The main task in petrophysical modelling is to generate areal variations in properties that attempt to reflect the geology and honour the well data. At the same time, it is usually required for the non-net facies, porosity and permeability to have similar trends, but with an element of heterogeneity. Vastly differing trends in the various properties will cause problems dynamically (e.g. high porosity, very low permeability). It is useful to compare the statistical distribution of the data for original well logs, blocked well logs and the 3-D model. There would normally be a close match between the original logs and blocked logs for each zone. However the 3-D model is designed to be dependant on well spacing, facies modelling, depth trends and it is NOT normally required for the statistics to be the same as the well data. It is possible to make comparisons within a particular area where properties are expected to match, such as within the channel belts of a reservoir, ignoring the poorer quality interchannel areas. Are the properties expected to be smooth are have a lot of variation between wells? How much dependence is there between net, porosity and permeability? Hydrocarbon Saturation In most reservoirs the hydrocarbon saturation can be readily calculated from the well logs and a hydrocarbon-water contact can be established from exploration wells with formation pressure tests (RFTs). Extending these observations to calculate the saturations in the rest of the reservoir requires further steps. Initial wells will obviously show original saturations, but as soon as production has started, subsequent wells will have higher contacts and flushed intervals so cannot be used for OOIP calculations. The hydrocarbon saturation is usually at a maximum in the crest of the structure and decreases significantly near the original oil-water contact. The oil saturation is also higher in good quality reservoir than poor quality reservoir. Therefore it is common practice to use a function that relates water saturation to height above the free water level to reservoir quality. Each cell in the model can then be evaluated in 3-D and water saturation calculated. Methods that should be considered are: Bulk Volume Water – Height (FOIL) does not need permeability Permeability – Height (e.g. J-function) relies on permeability calculation and special core analysis In both these method a geometrical property of height above the FWL needs to be generated in the model. While most methods rely on establishing a contact and this is important for static hydrocarbon volumes, it should also be recognised that the interval near the contact will not contribute very much to the reserves. It is not a good idea to artificially deepen the contact, based on very low oil saturations, unless there are special circumstances. Core fluorescence data may assist in defining the oil-water contact, but in some reservoirs there may be good fluorescence below the current free water level based on RFT measurements. Home Geological Modelling Purpose of Models Sources of Information Working Practices Steps in Geological Modelling Structural Definition Reservoir Layering & Facies Types Petrophysical Modelling Slideshow of Steps Dip in Subsurface Structure in Outcrop Measurement of Dips Scale of Dip Mapping Dips Dip in Cross Section Downloads Rescue Tools Geological Consulting