Second EAGE Workshop on Rock Physics Rock Physics: Integration & Beyond

12-14 January 2014

Muscat, Oman



RP09
Permeabty Predctons from Acoustc Core
Measurements
A. Draege* (Stato)
SUMMARY
Core measurements published by Weger et al. (2009) are used as a training dataset. We have used a
modiIied version oI the rock physics model Ior carbonates oI Sun (2000, 2004), to estimate permeability
solely as a Iunction oI P-wave velocity and porosity in brine saturated carbonates. The results were tested
against an independent database with 101 carbonate core measurements Irom seven diIIerent locations.
The test results are promising, indicating that over 72 oI the permeabilities can be predicted within one
order oI magnitude. An analysis oI the prediction errors indicated the validity areas oI the model. The
model struggles to reproduce permeability in high porous - low permeability rocks, which Irom thin
sections were Iound to be oomoldic. Also low porosity - high permeability carbonates were diIIicult to
predict. It is shown that integration oI the developed model with geological knowledge will improve the
prediction accuracy.
This methodology elucidates the intricate link between carbonate rock Irame stiIIness and pore structure
Ilow properties. It oIIers the potential to improve permeability predictions Irom wireline sonic data.


Second EAGE Workshop on Rock Physics Rock Physics: Integration & Beyond
12-14 January 2014
Muscat, Oman
Introduction

Carbonates are known for their highly varying and complex rock frames and pore structures that span
several size orders in scale. This makes the modelling of physical properties of carbonates
challenging. It is well known that the relation between permeability and porosity can be complex, but
studies like e.g. Baechle et al. (2008), Weger et al. (2009) and Xu & Payne (2009) have pointed out
that by including velocity in the permeability porosity relationship, a larger part of the permeability
observations can be explained. Velocity is related to permeability mainly through porosity and pore
shape. Generally high porosity entails lower velocities and high permeability. The carbonate pore
structure can influence these relations. Isolated rounded pores can for example result in high velocities
and low permeabilities in high porous carbonates. The effect of fractures can be opposite low
porosity fractured carbonates can have high permeability and low velocity.

This study is based on the database with carbonate core measurements published by Weger et al.
(2009), which is used as a training dataset. We have used a modified version of the rock physics
model for carbonates of Sun (2000, 2004), to estimate permeability solely as a function of P-wave
velocity and porosity. The results were tested against an independent database with 101 carbonate
core measurements from seven different locations (the blind test data set). The test results are
promising, indicating that over 72 % of the permeabilities can be predicted within one order of
magnitude. An analysis of the prediction errors indicated the validity areas of the model. This
methodology elucidates the intricate link between carbonate rock frame stiffness and pore structure
flow properties. It offers the potential to improve permeability predictions from wireline sonic data.

Method

Sun (2000, 2004) presented a rock physics model that accounted for various pore types in carbonates
through a parameter called frame flexibility factor ():

dry
= (1|)

(1)

The

is the frame flexibility factor for shear modulus (), and | is porosity. We can derive the
effective dry shear modulus from Vs and density of the rock: = Vs
2
. This can be used with
Equation 1 to estimate the frame flexibility factor for shear stresses. A similar relation can be used for
bulk modulus. In theory the frame flexibility is an important parameter in addition to porosity, when
estimating permeability. In this work we have modified the expression in Equation 1 to include P-
wave velocities (Vp) in brine saturated rocks instead of elastic moduli:
Vp = (1|)

Vp
(2)

This is done since Vp is a more common and less uncertain property compared to elastic moduli. We
also use the brine saturated samples directly, without the need for fluid substitution. The frame
flexibility factor for P-wave velocities can now be directly expressed by Vp and porosity:

) 1 ( log
) ( log
10
10
|

=
Vp
Vp
(3)

Permeability (K) can now be predicted by using the frame flexibility factor for Vp together with
porosity. A multivariate regression was performed to find four coefficients (a,b,c and d) to in the
following crossproduct:

d c b a
Vp Vp
+ + + = K ) ( log ) ( log ) ( log
10 10 10
| | (4)



Second EAGE Workshop on Rock Physics Rock Physics: Integration & Beyond
12-14 January 2014
Muscat, Oman
that minimized the error between predicted and measured permeabilities. We now have an empirical
model for predicting permeability directly from P-wave velocities and porosity.


Figure 1 Predicted versus measured helium core plug permeabilities on the Weger et al. (2009)
training data set. The white stippled line in the middle is where predictions coincide with
measurements (perfect fit), while the two other stippled lines are for deviations of one order of
magnitude. The A indicates how much of the data that are predicted with less than one order of
magnitude error. Thin sections of the cores marked with green dots are displayed in the upper part of
Figure 3.

Results

Calibration of the model to the training data shows a clear link between permeability, velocity and
porosity. Figure 1 shows predicted versus measured results of the calibrated data. The correlation
between predicted and measured data was high and more than 78 % of the data was predicted with an
error less than one order of magnitude.

The established relationship between permeability, Vp and porosity can be expressed as a rock
physics template, like shown in Figure 2. This template can be practical for visualization purposes,
and data can be plotted directly upon it to get a visual impression of the permeability. The velocity-
porosity values shown are calculated by Hashin-Shtrikman bounds, and added 200 m/s, to include a
broad range of Vp | combinations.

An implication of the model that can be seen in the template, is that permeability for a given porosity
value increases when Vp increases. This was confirmed from the training data (cf. Weger et al. 2009).
The reason for this is that cores with high amounts of micro pores lead to low velocities. Micro pores
also have a demolishing effect on permeability, hence the concurrent increase of velocity and
permeability as concentration of micro pores decrease.

Figure 3 shows thin sections of six samples from the training data. The upper three thin sections are
examples on where permeability predictions are good, even if permeability varies with more than 4
orders of magnitude. All the samples are from grainstones with interpaticle porosity. The low
permeability in the sample to the left is caused by high amounts of micro porosity. The two other
samples have low amounts of micro porosity. The three thin sections below show rocks with
approximately the same porosity. Sample 1 (left) is a grainstone with moldic pores and intermediate
amounts of micro pores, which together leads to low permeability despite relatively high porosity.
Sample 2 is a recrystallized dolomite with vuggy pores, and sample three is a framestone with vuggy
pores. Sample 2 and 3 have low amounts of micro pores. The model predicted too high permeability
for sample 1, while sample 2 was well predicted. The extremely high permeability caused by the
vuggy framestone in sample 3, was not possible to reproduce with the model.


Second EAGE Workshop on Rock Physics Rock Physics: Integration & Beyond
12-14 January 2014
Muscat, Oman
The model performance was further tested against the test data set, with 101 carbonate core plugs with
measured helium plug permeability equal or higher than 0.1 mD. Permeabilities lower than 0.1 mD
and higher than 4000 mD are set to be the model limits. As before permeability is modelled from P-
wave velocity and porosity. The results can be seen in Figure 4. The correlation coefficient is 0.48.
This is partly due to the high scatter at the lower limit of the model, 0.1 mD. More than 72 % of the
data are predicted within one order of magnitude deviation, which is encouraging. When analysing the
data in Figure 4, we find that the model digressed with reproducing low porosity high permeability
data, where permeability predictions were too low, see Figure 5. Also predicting high porosity low
permeability data was challenging, since permeability predictions were too high. By analysing the
reported pore types in each core, it became clear that the permeability prediction errors were related to
specific pore types; predictions of permeability in high porosity samples with oomoldic porosity are
almost exclusively too high, and predictions in low porosity samples with intercrystaline/inter particle
pores are generally too low. This is consistent with observations from the calibration dataset.


Figure 2 Rock physics template of Vp, porosity and permeability. The white stippled lines are iso-
permeability lines according to the model.


Figure 3 The upper three thin sections show examples where permeability predictions are good over
a range of more than 4 orders of magnitude. Permeability increases from left to right. The lower thin
sections show examples with approximately constant porosity, where the model 1) overpredicts
permeability 2) predicts permeability and 3) underpredicts permeability.


Second EAGE Workshop on Rock Physics Rock Physics: Integration & Beyond
12-14 January 2014
Muscat, Oman

Figure 4 Predicted versus measured permeability on the test data set. White stippled lines are as
defined in Figure 1.


Figure 5 Identification of permeability prediction on the test data set that was more than one order of
magnitude (OM) too low (magenta) and more than one order of magnitude too high (red). Green dots
were predicted with less than one order of magnitude deviation.

Conclusions

Permeability has been predicted from velocity and porosity by using an empirical approach. The
model is calibrated to a database, with more than 78 % of the predicted permeabilities deviating less
than one order of magnitude. A template for visual permeability analysis is constructed. When
evaluating the model for a test dataset of 101 core measurements, the model managed to reproduce
more than 72 % of the permeabilities within one order of magnitude. Analysis of pore types in the
different samples (Figure 3 and Figure 5) shows that there is a clear correlation between data with too
high predictions and pore types: Permeability in moldic and oomoldic pores are predicted to be higher
than measured, since they result in low permeability even at very high porosities. Close integration
with geological information can improve the predictability of the model. By excluding the few
measurements on oomoldic pores from Figure 4, there is hardly any data that are predicted with too
high permeability, and more than 78 % of the data are predicted within one order of magnitude.
Various pore types contribute in the carbonate cores with too low predicted permeability. The
common feature of these pores is that they manage to maintain a network, and maintain flow
properties even for low porosities. The velocity - permeability model is fast and simple, and a natural
next step will be to test it on well logs where also core measurements are available. Ultimately, the
presented model can potentially be used to identify flow units in a reservoir from seismic data.






Second EAGE Workshop on Rock Physics Rock Physics: Integration & Beyond
12-14 January 2014
Muscat, Oman
Acknowledgements

We would like to thank our colleagues Aart Jan Van Wijngarden and Giulio Cassini for discussions
and constructive contributions to this work, and Statoil for permission to publish this study. Further,
we thank the Comparative Sedimentology Laboratory at the University of Miami for data and
cooperation.

References

Baechle, G.T., Colpaert, A., Eberli, G.P. and Weger, R. [2008] Effects of microporosity on sonic
velocity in carbonate rocks. The Leading Edge, 27(8), 1012-1018. doi:10.1190/1.2967554

Sun, Y.F. [2000] Core-log-seismic integration in hemipelagic marine sediments on the eastern flank
of the Juan de Fuca Ridge. In: Fisher, A., Davis, E.E. and Escutia, C. (Eds.) Proc. ODP, Sci. Results,
168: College Station, TX (Ocean Drilling Program), 21-35.

Sun, Y.F. [2004] Seismic signatures of rock pore structure. Applied Geophysics, 1(1), 42-49.

Weger, R.J., Eberli, G.T., Baechle, G.P., Massaferro, J.-L. and Sun, Y.F. [2009] Quantification of
pore structure and its effect on sonic velocity and permeability in carbonates. AAPG Bulletin, 93,
1297-1317. doi:10 .1306/05270909001

Xu, S. and Payne, M.A. [2009] Modelling elastic properties in carbonate rocks. The Leading Edge,
28, 66-74.