SPE 105139
Fully Upscaled Saturation-Height Functions for Reservoir Modeling Based on
Thomeer’s Method for Analyzing Capillary Pressure Measurements
J.J.M. Buiting, Saudi Aramco
Copyright 2007, Society of Petroleum Engineers
This paper was prepared for presentation at the 15th SPE Middle East Oil & Gas Show and
Conference held in Bahrain International Exhibition Centre, Kingdom of Bahrain, 11–14 March
2007.
This paper was selected for presentation by an SPE Program Committee following review of
information contained in an abstract submitted by the author(s). Contents of the paper, as
presented, have not been reviewed by the Society of Petroleum Engineers and are subject to
correction by the author(s). The material, as presented, does not necessarily reflect any
position of the Society of Petroleum Engineers, its officers, or members. Papers presented at
SPE meetings are subject to publication review by Editorial Committees of the Society of
Petroleum Engineers. Electronic reproduction, distribution, or storage of any part of this paper
for commercial purposes without the written consent of the Society of Petroleum Engineers is
prohibited. Permission to reproduce in print is restricted to an abstract of not more than
300 words; illustrations may not be copied. The abstract must contain conspicuous
acknowledgment of where and by whom the paper was presented. Write Librarian, SPE, P.O.
Box 833836, Richardson, TX 75083-3836, U.S.A., fax 01-972-952-9435.
Abstract
Upscaling is a long standing problem for the construction of
saturation height models. Upscaling is difficult when desiring
to honor and retain the information obtained from core
descriptions, facies classifications and mercury injection
capillary pressure (MICP) data. In Saudi Aramco MICP
experiments are widely used for measuring the architecture of
the rock’s pore system. For the analysis of these measurements
Thomeer functions are standardly used. The Thomeer
parameterization proved to be very adept for handling the
complexity and multimodality of carbonate pore systems.
In this paper a direct and closed form solution of the upscaling
problem of MICP data to arbitrary large reservoir elements has
been derived for standard Thomeer functions. All statistical
variations of the essential rock properties, describing the
architecture of the pore-system, have been included in the
formalism. The upscaled result retains the familiar form of a
Thomeer function while replacing single core plug values with
their statistical equivalents. One of the important
consequences is that intrusion of liquids for large pieces of
rock can happen at much lower capillary pressures and thus
much closer to the free water levels in oil reservoirs.
Introduction
Within an oil reservoir the water saturation height functions
can vary strongly. In particular for carbonates these variations
can be significant and difficult to estimate. The amount of
hydrocarbons in a reservoir is the result of the balancing act
between buoyancy- and capillary forces. Capillarity is caused
by two physical phenomena, i.e. the liquid-rock interaction
and the architecture of the rock’s pore system. The latter
information is normally obtained through Mercury Injection
Capillary Pressure (MICP) experiments on core-plugs.
However, core plugs are tiny bits of rock from a specific part
of the reservoir and do not represent the properties of huge
piece of rock such as the grid cells in a reservoir model.
Significant errors can be the result in the transition zones
above the free water levels in carbonate reservoirs. The
thicknesses of these zones are often underestimated when
based on single or average core-plug parameters only, having
significant consequences for volumetrics and production
characteristics.
To calculate the saturations of large volumes of rock the full
statistics of their architectural parameters have to be taken into
account. For that it is required to upscale the information
obtained through MICP experiments from the tiny core plug to
the much larger reservoir dimensions. This upscaling has been
an ongoing challenge in the oil industry for many decades.
The two best known methods to analyze MICP measurements
and to estimate saturation height curves are Thomeer- and
Leverett J functions. However, neither method has been
satisfactorily upscaled until now. The popular J-function
method usual starts with a classification of pore systems using
the parameter κ φ . Ekrann1 has shown that saturation height
data parameterized in this way do not upscale. Additional
problems are encountered in carbonates with multimodal pore
systems, which are better handled using the Thomeer method2
for analysis of MICP measurements3.
In this paper a mathematically solution for upscaling Thomeer
functions to reservoir or well log dimensions is presented. The
derivations are supported by a wealth of core-plug data
available in Saudi Aramco. It has been found that the final
upscaled result retains the familiar form of a Thomeer function
while replacing the elemental core-plug values of each specific
petrophysical rock type with their statistical equivalents. In
other words the statistical variations of the fundamental rock
parameters within the larger rock elements play an integral
part in the behavior of the upscaled saturation-height
functions. As a result, hydrocarbon saturations in critical parts
of the reservoirs, such as the transition zones, can be much
more realistically modeled. For example, the upscaling
exercise demonstrates that transition zone thicknesses,
determined from only average core-plug data, underestimate
the transition zone, which could have significant effect on the
estimation of the oil in place.
2 SPE 105139

Some preliminary remarks: mercury at any Pc is than given by

• In this paper a solution will be derived for reservoir
rock with a monomodal pore system only. It is
believed that the generalization to multimodality
Bv (Pc ) ≈ φ ⋅ exp ( −G
log( Pc ) − log( Pd ) ) for Pc> Pd (2)
would add additional complications to a paper that is
not simple already and that it is better to leave it out Please note that the exp() term in this equation is in fact the
of here. Carbonates, though, are in general bimodal or mercury saturation of the core-plug, i.e.
even trimodal. The extension to multimodality of the
results as presented here is relatively straightforward.
S Hg (Pc ) ≈ exp ( −G
log( Pc ) − log( Pd ) )
• The distinctive shape of the Thomeer functions are
The conversion from SHg to Soil is normally done via the
the result of astute observations of Mr. Thomeer, who
interfacial tension values σ·cos(θ) for crude, brine, rock and
noted that the relation between the experimental
Hg, air, rock.
log(Bv) values and the applied pressures log(Pc)
exhibit a hyperbolic relationship. He translated this in
an empirical formula, which works very well in
describing MICP experiments. However, the FIG.1: Typical Thomeer curve
expression has several mathematical idiosyncrasies,
250
which made the upscaling exercise not a simple. I
hope the reader bears with me.
200

1. The basics
15 0
From experimental data Thomeer observed hyperbolic relation
between the amount of intruded mercury and the applied Pc
pressure in the log-log displays. Based on these observations 10 0

he proposed an empirical relationship1:

( )
50
⎧⎪B ⋅ exp −G
for Pc > Pd
Bv (Pc ) ≈ ⎨ v ,∞ log( Pc ) − log( Pd )
(1)
⎪⎩ 0 elsewhere 0 Pd
20 Bv, ∞ 15 10 5 0
Bv,∞ - Fract. Bulk Vol. occupied by Hg at Pc=∞
G - Pore Geometrical Factor (G <2 in general) Bv, occ
Pc - Capillary pressure (applied Hg pressure)
Pd - Minimum entry pressure 2. Q-domain representation
Even though the Thomeer parameterization is very successful
This functional form appeared to be both practical and and widely applied, the corresponding function – as seen in
successful. Many companies use the Thomeer Eq.1 - has some peculiar mathematical idiosyncrasies, which
parameterization for MICP experiments and for describing the are very tricky to deal with as will be seen in this paper. To
internal architecture of a reservoir rock. Also in Saudi Aramco being to deal with these difficulties and thus for the sake of
Thomeer parameters and functions are standardly used. An mathematical simplicity all derivations will be done in the “Q-
example of a typical Thomeer hyperbola is shown in Fig.1. domain”, which is logarithmic equivalent of the pressure
domain, i.e.
The work presented in this paper is on the upscaling of these Q = ln(Pc ); Qd = ln(Pd ) and g = 2.3G
Thomeer functions. It is also very much instigated by the
issues occurring in our carbonate reservoirs, i.e. the Eq.2 then obtains the form of:
derivations focus strongly on carbonates and it is important to
remember that carbonates have in general multiple pore ⎧ −g
φ , g , Qd ⎪ ≈ φ ⋅ e Q − Qd for Q > Qd
systems. For the sake of simplicity we treat in this paper Bv (Q) ⎨ (3)
mono-modal pore systems only. The generalization, however, ⎪⎩= 0 elsewhere
to a bi- or tri-modal pore system is straightforward.
In this paper it will be shown that the Thomeer hyperbolae of
One more simplification will be introduced: for a carbonate Eq.2, as matched against the individual MICP core plug
with a mono-modal pore system it will be assumed measurements can be up-scaled to larger reservoir dimensions
that Bv ,∞ ≈ φ *, i.e. the fractional bulk volume occupied by and still retain certain relationships to the core plug Thomeer
parameter distributions. This upscaled form can be written as
*
This is not fully true and Bv,∞ tends to be slightly larger than φ for
larger bulk porosity values. Also it has to be kept in mind that we
consider here a mono-modal pore system.
SPE 105139 3

−g
 which states that only those plugs with minimum-entry-
Bvup (Q ) ≈ φ ⋅ e Q −Qd , pressures Pd less than the applied pressure Pc will have been
penetrated by the mercury. Using Eq.3 this expression can be
which is similar to the Thomeer form, as seen in Eq.3? The simplified to:
plug related porosity φ and curvature exponent g, are replaced
−gi
by their respective average values, i.e. φ * and g , which are 1 N

∑
Q −Qd , i
Bvup (Q ) = φi ⋅ e (6)
constants. The minimum-entry-pressure terms, Qd or Pd, are N
  i =1
Qd , i <Q
replaced by Qd (or Pd ), which – however - is not a constant
but a function of the capillary pressure Q or Pc, as will be seen
This is the basic expression to be addressed, and already
later.
something can be said about its asymptotic behavior, i.e.
1 N
3. Upscaling problem
The Thomeer parameters of a tiny core plug cut out of a large
Lim Bvup (Pc ) = Lim Bvup (Q ) =
Pc →∞ Q →∞
∑ φi = φ
N i =1
(7a)

piece of rock are as such not representative for its internal Lim Pc = Pd ,min . or Lim Q = Q d ,min . (7b)
architecture. A typical plug has a diameter of 1.1cm and a Bvup → 0 Bvup → 0

length of 2.54cm, i.e. a volume of about 9.6 cm3. A reservoir
rock element, such as the ones probed by wire-line logs,
represents a volume sample of at least 50,000 of these core
plugs in size. The dimensions of a typical grid cell in a
reservoir model is in the order of 250m x 250m x 1m, which
would correspond to an equivalent of billions of core plugs
and it difficult to defend that the parameters derived from a
few core plugs could represent the capillary properties of such
a big piece of rock.
Qd,min is the related to the lowest possible minimum entry
pressure, i.e. is located at the left edge of the (normal)
distribution function.
Since N is large, the summation of Eq.4 can be replaced by an
integral, i.e.
Bvup (Q) = ∫∫∫ Ω(φ , g , Qd ) Bvφ , g ,Qd (Q) dφ dg dQd (8)

A large rock volume can be seen as a vast population of core where Ω(φ , g , Qd ) is the 3-dimensional distribution
plugs, each with their fundamental rock properties in terms centralized on the average position (φ , g , Q ) , with the d

of φ, G and Pd. These parameters will be distributed around condition

their average values, valid for the rock volume. A study of
over five hundred core plugs from the Ghawar Arab D
limestone, on which MICP and Thomeer type curve matching ∫∫∫ Ω(φ , g , Q ) dφ dg dQ d d =1 (9)
was performed, has found that φ and G are overall normally
distributed. The Pd ‘s exhibit a log-normal type of distribution,
which implies that Qd = ln(Pd) tend to be normally
distributed†. Fig.2 shows a suite of Thomeer curves from a
large φ, G, Pd distribution. Also the average Thomeer function
is shown. The spread between all these functions is large.

F IG .2: D is trib u tio n o f T h o m e er C a p -c u rve s

25

If N is the number of plugs in a rock element, then the

percentage bulk volume occupied by mercury of the upscaled
20

rock element will be:

N
1
Bvup (Pc ) = ∑ B (P ) = B (P ) . (4)
15

T h o m eer
averag e-

v ,i c v c
B v (P c )

N i =1

The upscaled Bv at a certain pressure Pc is determined by the

10

averaging of the entire plug Bv’s at that Pc. Inserting the

expression for of Eq.1 yields the following conditional
summation:
5

( ),
N
1
Bvup (Pc ) =
N
∑ i =1
φi ⋅ exp −G i
log( Pc ) − log( Pd , i )
(5)
Pd , i < Pc
0

Pd_av er.
0

50

100

150

200
Pc

*
In principle this should read B v , ∞ .
†
For coupled normal and lognormal distributions the following
relation holds: Q d = ln( Pd , median ) .
4 SPE 105139

Eq.8 is a weighted averaging of all the plugs in the rock
element for a certain pressure Q (assuming that Q>Qd,min). The
result is the expectation value of all the Bv’s, which will be a
certain Bv of the rock element at location (φ , g, Q), i.e.
Bvup (Q ) = E [ Bv ] = Bv φ , g ,Qd
(Q ) = φ ⋅ e
Note that Gaussians have unlimited ranges. Practically we
circumvent this by relating the standard deviation to the
minimum and maximum (or low and high) values of the
parameters, i.e. σ ≈ 16 (x max − x min ) and it is assumed that the
− g
values outside the range [x − 3σ , x + 3σ ] are negligible§. Eq.8
Q −Qd
(10) can than be written as:

φmac g max Q
All three parameters φ , g and Q are in principle dependent on
−g

∫ ∫ ∫
σφ σg
Bvup (Q ) ≈ fφ (φ ) f g ( g ) fQσ (Qd ) φ e Q − Qd
d φ dg dQd (15)
Q (i.e. of the capillary pressure Pc ) and – as indicated by the φmin g min Qd ,min
d

˜symbol - related to their respective average values. As a

matter of fact φ = φ and g = g , as will be seen later. This is the basic integral to be solved.

From the numerous carbonate measurements available it was

observed that in general the correlations between φ, g and Qd
are weak. Only for high entry pressures G and Pd are stronger 4. The φ-integral
correlated, but that is only for a small portion of the core plug The porosity part of the integral can be taken outside the
population.* Because of the weak inter-parameter correlations combined integration:
Ω can be approximated as the multiplication of the three
independent centralized distributions: φmac
σ
g max Q
σ
−g

∫ ∫ ∫
σg
Bvup (Q ) ≈ fφ φ (φ ) φ d φ fφ φ f g ( g ) fQσ (Qd ) e Q − Qd
dg dQd (16)
 
g d

Ω(φ , g , Qd ) ≈ ψ φ (φ ) ⋅ψ g ( g ) ⋅ψ Qd (Qd ) (11) φmin

=φ
min Qd ,min

or
g max Q
For the measured carbonates samples, it can be seen that the −g

∫ ∫
σ
Bvup (Q ) ≈ φ f g g ( g )f Qσd (Qd ) e Q −Qd dg dQ d (17)
pore geometrical factors G’s are more-or-less normal g min Qd ,min
distributed. The minimum entry pressures Pd – which are
directly related to the pore-throat radii – show a skewed
distribution curve, resembling a log-normal distribution. As a This resolves the φ (or Bv,∞) contribution to integral Eq.15.
result, the Qd’s are approximately normal distributed†. It is The averaged result as expressed in Eq.10 can be written as:
assumed that the φ ' s are also normally distributed, but that is
− g
in fact irrelevant as will become clear later on. As a result the up
B (Q ) = Bv φ , g ,Qd
(Q ) = φ ⋅ e Q −Qd
, i .e . φ = φ (18)
v
following normal distribution is used for all three, Q-domain,
Thomeer parameters:
Moreover, the up-scaled Bv is independent **of the variation in
−1 ( x − x )2 porosity, i.e. σφ ... Note that the above result is valid for any
ψ x ( x) ≈ f xσ ( x) = σ 1
2π
e 2σ 2 (12) shape of the porosity distribution.

x Is the mean and σ the standard deviation. These are

normalized Gaussians and Eq.11 can be expressed as follows:

σφ σ 5. The g-integral
Ψ (φ , g , Q d ) ≈ f (φ ) ⋅ f g g ( g ) ⋅ f Qσ (Qd ) ‡ (13) Rearranging the terms in Eq.17 yields
φ d

Of course normalization condition holds: Q ⎡ g max σ g −g ⎤

∫ d ⎢ ∫ f g (g ) ⋅ e
σ
Bvup (Q ) ≈ φ f Qd
(Q ) Q −Qd
dg ⎥ dQ d (19)
Qd ,min ⎣⎢ g min ⎥⎦
∫∫∫f
σφ σ
φ
(φ ) ⋅ f g g ( g ) ⋅ f Qσd (Qd ) d φ dg dQ d = 1 (14)
Here we introduce another simplification by removing the φ-
contribution altogether for the time being, i.e.

* g
Please note that this holds for carbonates. For clastics significant −

correlations do exist. The effects intra-parameter correlations will be bvup (Q) ≡

Bvup ( Q )
φ
=e Q − Q d
(Q > Q d ,min ) (20)††
treated in a later paper.
†
Moreover it was observed that for carbonates Pd >1 psi and thus §
About 1% of the Gaussian distribution falls outside the range.
that Qd >0. **
‡ For weak correlations between the parameters!
Note that for reasons of convenience and readability the Qd †† 

subscript for the standard deviation σ in the above and following This actually is the Hg-saturation, i.e. S Hg (Q ) = exp( Q −−Qg ) in case of
d

formulas is omitted, i.e. σ ≡ σ Qd a mono-modal system and Bv ,∞ ≈φ .

SPE 105139 5

The integral of Eq.19 then becomes: +

σ g2
2(Q −Qd )2
Note that lim e = ∞ ,i.e. this exponential is divergent,
Qd →Q
Q
⎡g σ max −g ⎤
bvup (Q ) ≈ ∫ f Q (Q d ) ⎢ ∫ f g ( g ) e but also observe that this singularity is annihilated by the
σ g Q −Q d
dg ⎥ dQ d (21)
⎣⎢ g ⎦⎥
d
Q d ,min min second term in Eq.23, i.e.

The first integral in Eq.21 to tackle is the g-integral: σg σg

⎧⎪ ≈ 1 for Qd < Q − σ g
3
Λ g ' ( g max ) − Λ g ' ( g min ) ⎨ σg
.
g max −g
g max −
( g − g )2
−g
⎪⎩≈ 0 for Q − 3 < Qd < Q
∫ ∫
σg 2σ g 2
Ig = f g (g )e Q −Qd
dg = 1
e e Q −Qd dg (22) σ g2
σg 2π where g ' = g + ( Q −Qd ) !
g min g min

With this observation Eq. 23 can be expressed as:

After some rearranging this can be rewritten as:
2 ⎧ −g σ g2 σg
−g
+
σ g2 g max −
1 ⎧⎪ σ g2 ⎫ ⎪
⎨ g − g + (Q −Qd ) ⎬ ⎪⎪e (Q −Qd )
+
2(Q −Qd ) 2 for Qd < Q − 3
2σ g 2 ⎩⎪
∫
(Q −Qd ) 2(Q −Qd )2 ⎭⎪
Ig =e σ g 2π
1
e dg . Ig ≈ ⎨ (25)

 

g min ⎪ σg
σ σ σ g2 ⎪⎩ 0 for Q − 3
< Qd < Q
=Λ g g' ( g max ) −Λ g g' ( g min ), where g ' = g + (Q −Q
d)

σ g2
This converts the Ig integral into: σg +
2 (Q −Qd )2
For Qd < Q − 3
the e term is well behaved and, in
−g σ g2 essence, introduces only a small, constant shift ∆g in the Q-
+
Ig =e (Q −Qd ) 2(Q −Qd )2
{Λ σg
g' ( g max ) − Λ g ' ( g min )
σg
} (23) domain as can be seen in Fig.3.
σ2
This shift can be estimated for Qd < Q − 2 gg (which holds
Where Λσx (x ) is the probability function defined as since Qd < Q −
σg
and g ≈ 3σ g ) as follows:
3
2
x '= x ( x '− x )
−

∫
2σ 2
Λσx (x ) = σ 1
2π
e dx ' (24) −g σ g2 −g ⎧⎪ σg 2 ⎫⎪
+ ⎨1− ⎬
−∞ 2 (Q −Qd ) ⎪⎩ 2 g (Q −Qd )⎪
The individual components are displayed in Fig.3. e (Q −Qd ) 2(Q −Qd )
=e ⎭
(26)
−g
1.2 80 {1−σ 2
g 2 g (Q −Qd ) }
= e (Q −Qd )
1.0

Since σ g2 2 g (Q − Q d )  1 the above equation can be

0.8 approximated as:
blue curve

0.6 −g σ g2 −g 1
+ ⋅
(Q −Qd ) 2(Q −Qd ) 2 (Q −Qd ) 1+σ g2 2 g (Q −Qd )
e ≈e or
0.4

−g σ g2 σ g2 −g
+ − g (Q + −Qd )
∆g (Q −Qd ) 2(Q −Qd )2 Q + ∆ g −Qd
0.2 e ≈e 2g
=e (27)
Q
σ g2
0.0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
0
From which follows that: ∆ g ≈ 2g (28)*
Qd=ln(Pd)
Since ∆g >0 and Qd < Q, the integral Ig of Eq.25 can be
Fig.3: The components contributing to the g integration. The multiplication by the
positive and singular exponential (blue curve) causes a shift ∆g in the Q-doamin.
expressed as:

⎧ (Q +∆− g−Q ) for Q < Q

⎪e g d d
Ig ≈ ⎨ (29)
⎪
⎩ 0 elsewhere

So in conclusion, the upscaling step in the g-domain reduces

the distribution of g’s to its average value, i.e. g = g , and, as

*
A typical value is ∆g≈0.07.
6 SPE 105139

an additional effect, introduces a small and positive shift in no straightforward indefinite solution is available‡. To
the Q-domain, i.e. Q → Q + ∆ g *.As a result Eqs.10 and 18 circumvent this problem some pragmatic approximations had
evolve into: to be found.

Fig.4a: Thomeer functions and the Q-distribution function

−g
1
φ , g ,Qd Q +∆ g −Qd
v
up
B (Q ) = Bv (Q ) = φ ⋅ e . (30)

So in effect φ = φ and g = g , with the caveat that the

variation in the g-domain, i.e. σg, remains a factor through the 0.5
shift ∆g in the Q-domain.

6. The Qd integral: preliminaries 0

Through the result of Eq.29, the integral of Eq.21 reduces to: -1 0 1 2 3 4 5

Q d -ln(P d )

Q −g
Fig.4b:
∫
(Q +∆ g −Qd )
bvup (Q ) ≈ f Qσd (Q d ) e dQd (31)
Qd ,min

0.7
σ
Since f Qd (Q d ) ≈ 0 for Qd < Q d ,min , the lower range of this 0.6

†
integral can be extended from Qd,min down to -∞ , i.e. 0.5

0.4
Q −g

∫f
σ (Q + ∆ g −Qd ) 0.3
bvup (Q ) ≈ Qd
(Q d ) e dQ d
−∞ 0.2

or 0.1
Q g (Qd −Qd )2
− −

∫e
(Q +∆ g −Qd )
bvup (Q ) ≈ σ 0.0
2
1
2π
e 2σ
dQd (32)
0 1 2 3 Q d4=ln(P d ) 5
−∞

Fig.4: Comparing the individual components comprising the integral of Eq.32

The behavior of the two exponential functions in Qd-domain is in Fig4a and the effect of the multiplication in Fig.4b. The multiplication
displayed in Fig.4a, keeping in mind that the Thomeer results in distorted bell curves. The original normal distribution of the Qd’s is
shown as a dashed curved, enveloping all the other ones. The integrals of
function are zero for Qd>Q. The combination of the Eq.32 represent the areas under the distorted bell curves of Fig.4b and are
exponentials is shown in Fig.4b and the integral of Eq.32 well behaved, disregarding the improperness of the Thomeer function.
represents in effect the areas under the distorted bell curves as
seen in Fig.4b.
−g
(Q +∆ g −Qd )
Eq.32 is in fact the weighted average of e and
therefore
6.1 Solving via a coordinate transformation
g
Because of the improperness of the integral of Eq.33 a
−
(Q +∆ g −Qd ) pragmatic solution will be derived by proxy. To achieve this
bvup (Q ) ≈ e (33)
coordinate transformation centered around the point Q+∆g is
introduced:
where Qd,min < Q̃d < Q .
y = g1 (Q + ∆ g − Qd ) - related to min. entry press.
There are two difficulties with integral Eq.33. Firstly its range
is limited to Qd ≤ Q. In other words, only plugs with a y = g1 (Q + ∆ g − Qd ) - related to aver. min. entry press.
minimum entry pressures Pd less than the applied pressure Pc (34)
y = g1 (Q + ∆ g − Qd ) - the upscaled version
are being filled with mercury and contribute to the total
amount of liquid in the rock element. Secondly the function y o = g1 (Q + ∆ g − Qo ) - for later usage
has a singularity at Qd = Q+∆g and even though the
integration does not reach that far, the integral is improper and
‡
However since the roots are outside the integration limits (∆g is a
* positive number) a finite value exists, as can be concluded from Fig.
Or a negative shift in the Qd-domain: , i.e. Q d → Q d − ∆ g . 4b as well.
†
For carbonates in general Pd >1 psi, i.e. Qd >0.
SPE 105139 7

Moreover, δ y =
∆g
(=
σ g2
> 0) and σ y = σg . 6.2 High pressure solution
g 2g 2 For large values of the applied mercury pressures, i.e. Pc ‘s
above the highest possible minimum entry pressures Pd,max , all
Note that since the pore geometrical factors G (and thus g) are elemental rock elements (core-plug equivalents) are being
normally distributed and positive by definition, the maximum intruded by mercury. This implies that Pc > Pd,max and thus
size of its standard deviation is limited by the value of its Q > Q d ,max ≈ Qd + 3σ . In terms of the y’s this means (see
average value, i.e. σ g ≤ 13 g → δ y ≤ 181 . Moreover Q > Q d Eq.34) that
or y > δ y * and Eq.32 can be expressed in the y-domain as: y > 3σ y + δ y (37)

∞ −
1 −
( y − y )2 and that the range of the Gaussians or bell curves fall totally
2σ y2
∫ within the validity range of β(y), i.e. y > 0, as can be seen by
up
b (Q ) =
v σy
1
2π
e y
e dy (35)
δy Fig.5. Note that by definition β(y<0)=0. Moreover,
β(0<y<δy)≈ 0 – as is also explained in the Appendix - and that
−1 therefore the lower integration limit of Eq.36 can be extended
Introducing the function β ( y ) = e y for y > 0 from δy to -∞, i.e.
and β ( y ≤ 0) = 0 , Eq.35 expressed in terms of this β(y)
( y − y )2
becomes: ∞ −
dy = E [ β ] = β
2σ y2
∞ −
(y −y )
2σ y2
2
Iy = 1
σ y 2π ∫ β (y ) e (38)
bvup (Q ) = σy
1
2π ∫
δy
β ( y )e dy ≡ I y (36) −∞

This extension has the advantage that the standard definitions

Some of the properties of this function are summarized in the
for the statistical parameters of a normal distribution such as
Appendix. Note that β(y) is analogous to Eq.20. Moreover
mean and variance can be employed. In particular for
since 0 < δy < 0.06, and thus small, β (0<y<δy)≈ 0 for all
functions which vary slowly compared to the standard
practical purposes and it is possible to extend the lower limit
deviation, such as is the case for the high pressure bell curve
of the above integral to 0.
in Fig.5, where the β-function is more-or-less linear over the
range of the bell-curve. In these cases the β-function in the
FIG.5: Comparing β (y) with gaussians related to three
pressure regimes vicinity of y can be approximated by a Taylor expansion

1
<y>=0.9
(med Press)
<y>=3 β ( y ) ≈ β ( y ) + ( y − y ) β ' ( y ) + 12 ( y − y ) 2 β '' ( y )
<y>=-0.2 (high Press)
(low Press)

β(y)=exp(-1/y)
yielding the following result for the integral:

( y − y )2
∞ −
2σ y2
σy
1
2π ∫
−∞
β (y ) e dy ≈ β ( y ) + 12 σ y2 ⋅ β '' ( y ) (39)

∂2 β ⎛ 1 2 ⎞
-1.5 -1 -0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
where β '' ( y ) = = β (y )⎜ 4 − 3 ⎟ (40)
∂y 2 ⎝y y ⎠
δy y y =y

and thus
Fig.5: The β(y) function and 3 normal distributions related to different pressure
I y ≈ β ( y ) ⎡1 + ⋅ ( y14 − y23 ) ⎤ for y > 3σ y + δ y
σ y2
regimes. Only the high pressure curve falls within range of y > δy .The other two (41)
only have partial coverage of β(y) and the integral Iy is more difficult. ⎣⎢ 2
⎦⎥

In Fig.5 β(y) and three normal distributions are shown, each Since β(y) varies slowly for high y values the Iy is the
related to a different pressure†. For the high pressure case the expectation or mean value of the population of the β’s. The
Gaussian covers a small, almost linear part of the β(y)-curve result should be a β-value closely located to the average of the
and the resulting mean value β ( y ) ≈ β ( y ) . In the low parameter y, i.e. y :
pressure case, only values between ≈ 0.2 and 1.0 contribute to
the integral of Eq.36. I y = E [ β ] = exp ( −1
y + δ hp ) = β ( y ), where y = y + δ hp (42)

*
Note that even though y is always positive, y can be negative. For small δ hp a Taylor expansion can be applied:
†
Here it needs to be noted that the values of the y ’s not only depend
δ hp
on the pressures Q and Qd, but also on the values of the g’s (Eq.34), β ( y + δ hp ) ≈ β ( y ) + δ hp ⋅ β '( y ) = β ( y ) ⋅ (1 + y2
) (43)
i.e. high y values can also be a sign of an abnormal low g value.
8 SPE 105139

Combining these results with Eq.41 yields: ⎡ −g ⎤

Bvup (Q ) = φ exp ⎢ ⎥, (46)
⎣⎢ Q + ∆ g − Q d − ∆ hp ⎦⎥
β ( y ) ⋅ ⎡⎢1 + ⋅ ( y14 − y23 ) ⎤ ≈ β ( y ) (1 +
σ y2 δ hp
) (44)
⎣ 2 ⎥⎦ y2

As mentioned: ∆ hp = − g δ hp and is positive and well behaved

From this equation δhp can be solved, i.e. for high pressures and/or low g values, i.e. in the region
σ y2
where y > 3σ y + δ y in Fig.6. From Fig.6 it can also be seen
δ hp ≈ ⋅ ( y12 − y2 ) for y > 3σ y + δ y (45)
2 that this shift becomes anomalous for low y values. To avoid
this problem a low- y tapering is introduced, as already
indicated in Fig.6. This results in the following high-pressure
0.7 shift function in the Q-domain, which is positive for all values
of Q and has no effect for low Q-values:
δ hp - Eq.45

⎧ σw (1 − 2gw )
0.5 2
for w ≥ w o
⎪
Hi-press. Shift δ hp

0.3
∆ hp ≈ ⎨ σ 2 g
⎪⎩ w o (1 − 2w o ) exp 2σ hp2 (w − w o )
−1
(
2
) for w < w o
(47)

where
3σ y+δ y w = Q − Qd + ∆ g
0.1
wo =σ +g
-2 -1 0 1 2 3 4 5 6 7 σ hp ≈ 13 (Qd + σ + g )
-0.1
tapering

The parameters wo and σhp are chosen in such way to assure

-0.3
Fig.6: The high pressure shift δhp of Eq.45. For high pressures, i.e. beyond the
point 3σy + δy the shift is a slowly varying negative number. Before that point
the function becomes anomalous. To avoid this behavior a tapering can be
applied as is shown by the red curve.
The behavior of δhp as function of y can be seen in Fig.6.
Note that δhp=0 for y = 1 2 and becoming singular at y = 0 .
Within the validity range for high pressures, i.e. for
proper tapering to the low Q-side and sufficient shifting
effects for the higher Q-values. This function is shown in
Fig.7 for several values of the average (log) minimum entry
pressures and average pore geometrical factors, i.e. several
(Qd , g ) combinations.
0.3
Qd=2,g=0.6

y > 3σ y + δ y (see Eq.37) the shift δhp is negative and slowly Qd=2,g=1.0
Qd=2,g=2.0
approaching 0 in an almost linear fashion. This shift results
from the second-derivative of β (Eq.39), yielding the variance
hi-press. Q -shift: ∆ hp

0.2
dependency as seen in Eq.45. It will be used later on.
For smaller values of y the δhp function of Eq.45 is not well
behaved. In these cases the high y assumptions, embedded in
Eqs. 38 and 39, are not valid anymore. In Fig.6 a tapering (red
curve) is applied for those values. The reasons will be 0.1
explained below.

With the expression for the δhp shift, valid for high y values,
the high pressure (and/or low g ) case of the upscaling 0
problem has been solved. From Eq.42 it is known that 0 2 4 6 8
y = y + δ hp . Using the coordinate transformations of Eq.34 Q=ln(Pc)

the upscaled (log) minimum entry pressure function Qd can

Fig.7: Examples of the he high pressure shift function ∆hp
be written as Q = Q + ∆ . Here ∆ = − g δ , which is a
d d hp hp hp as defined by Eq.47 for three different combination of
positive number for the high pressure regions. Substituting this (Q d , g ) . Note the low Q tapering. For high Q’s the shift
into the intermediate result of Eq.30, yields the high pressure goes to zero.
solution:
SPE 105139 9

6.3 Low to medium pressure regime ∞ −

( y − y )2
2σ y2
For Q < Q d + 3σ = Qd ,max , the integration over Qd , and thus y ≈ y o − y o 1
σ y 2π ∫e
yo
dy
over y (Eq.36), can not be extended anymore over the range -∞
( y − y )2
→ +∞ and a different approach to solve the integral is needed. ∞ −
2σ y2
Again the problem will be tackled using the y-transformation +y σy
1
2π ∫e dy (52)
defined by Eqs.34 and 35, i.e. by solving: yo

( y − y )2
∞ −
2σ y2
∞ −
1 −
( y − y )2
2σ y2
+σ 1
2π ∫ ( y − y )e dy
π ∫
Iy = e y
e dy = β ( y ) (48)
1 y
y =yo
σy 2
δ y

The first two integrals are complementary probability

A graphical representation of the contributing functions is functions (compare to Eq.24). The 3rd integral can be solved
shown in Fig.5, but now we concentrate on the bell curves by changing the integration variable from y to ( y − y ) , i.e.
related to the low- and medium pressures.
y ≈ y o − y o ⎡⎣1 − Λ y y ( y o ) ⎤⎦ + y ⎡⎣1 − Λ y y ( y o ) ⎤⎦
σ σ
As argued before, Eq.48 yields an average value of the β’s
with respect to a normally distributed y’s: ∞ −
( y − y )2
2σ y2

∞
(y −y ) 2
+σ
y
1
2π ∫
y −y =yo −y
( y − y )e d (y − y )
−
2σ y2
Iy = σy
1
2π ∫ β (y ) e dy = β = β ( y ) (49)
δy where Λ is the probability function (see Eq.24). Note that the
expression [1-Λ] is equivalent to the complementary
where ỹ is assumed to be somewhere between δy and y+3σy. probability function.
The objective is to find an accurate estimate of this ỹ. This is Further rearrangements lead to:
achieved by using the fact that for smaller y’s, and for the ∞ −
z2
2σ y2
y ≈ y o ⋅ Λ y ( y o ) + y ⎡⎣1 − Λ y ( y o ) ⎤⎦ + σ 1 2π ∫
σy σy
purpose of finding a good ỹ, a linear approximation is made, 1
2 e dz 2
i.e. y
z =yo −y

z2
By substituting u = one gets
⎧ − y1 2σ y2
⎪
β ( y ) = ⎨e ≈ α ( y − y o ) for y o < y < y + 3σ y (50) 2σ 2
∞
y ≈ y o Λ y y ( y o ) + y ⎡⎣1 − Λ y y ( y o ) ⎤⎦ + 2y ∫
σ σ
⎪⎩0 for y < yo σy
1
2π
e −u du
2
(yo −y )
u=
2σ y2
The line α(y-yo) passes to the inflection point of the β-function And thus
at y = 12 , where β ' ( y ) is at its maximum and thus β '' ( 12 ) = 0 . ( y o − y )2
−
2σ y2
y ≈ y o ⋅ Λ y ( y o ) + y ⎡⎣1 − Λ y ( y o ) ⎤⎦ + σ
σy σy 2
Then the slope α = β ( ) = 0.54 and it intercepts the y-axis at
' 1
2
1
y σ y 2π e
y o = 14 . See also the Appendix. Substituting the linear
approximation in Eq.49 yields: and via Eq.12:

y ≈ y o ⋅ Λ y y ( y o ) + y ⎡⎣1 − Λ y y ( y o ) ⎤⎦ + σ y2 ⋅ f y y ( y o )
σ σ σ
∞ −
( y − y )2 (53)
2σ y2
σy
1
2π ∫ α (y − y
yo
o )e dy ≈ α ( y − y o ) (51)
In general Λσx (x ) = Λσ0 ( x − x ) and f xσ (x ) = f 0σ ( x − x ) and
Eq.53 can be written as (omitting the “0”-subscripts):
From which ỹ can be solved, with the understanding that it is
good for those y ranges where the linear approximation yields
y ≈ y o ⋅ Λ y ( y o − y ) + y ⎡⎣1 − Λ y ( y o − y ) ⎤⎦ + σ y2 ⋅ f
σ σ σy
a ỹ comparable to the one we would get by solving Eq.49 for (y o − y )
β(y). This works amazingly well for the medium pressure
cases, but some adjustments are needed for the very low Resulting in the medium pressure solution:
pressures and of course for the high pressures.
σ σy
Note that the slope α drops out of eq.51 and we get: y ≈ y + ( y o − y ) Λ y ( y o − y ) + σ y2 f (y o − y ) (54)
2 2
(y −y ) (y −y )
∞ − ∞ −
2σ y2 2σ y2
Which is valid for y < 3σ y + δ y , but not for the lowest
y ≈ y o − y o σy
1
2π ∫ e dy + σ
y
1
2π ∫ ye dy
yo yo
values of y . This is caused by the fact that for y < 0.211 (see
By replacing y in the second integral with ( y − y ) + y one Appendix) β(y) suddenly flattens out and almost becomes
obtains: zero. The linear approximation does not hold for this situation
10 SPE 105139

anymore, since the line crosses the y-axis at y=¼, i.e. beyond σ
( y − y − δ hp ) ≈ ( y o − y − δ hp ) Λ y ( y o − y − δ hp )
the 0.211 point. σy
+ σ y2 f ( y o − y − δ hp ) (57)
σy
− y o ⋅ Λ ( y o − y − δ hp − δ xx )
6.4 Amending the lowest y values
Eq.54 represents an accurate solution for medium values of where δ xx = δ lp − δ hp
y , but not for very low y values, i.e. y < −2σ y when
Eq.57 is a very symmetric expression† and applies to all valid
Λ ( y o − y ) ≈ 1 and f ( y o − y ) ≈ 0 and Eq.54 yields that
values of y , i.e. low, medium and high. It represents the
y ≈ y o . This is not correct, since in these circumstances y is
solution of the upscaling problem in the y-space, i.e.
will be smaller y o and should approach 0. Therefore a small
correction has to be subtracted from Eq.54. This low y or
Bvup (Q) = φ exp − 1y( )
However it still needs to be transformed back to the Q-
low-pressure correction can be expressed as:
domain.
σ *
y o ⋅ Λ y ( y o − y − δ lp ) , where δ lp ≈ 2.7σ y (55)

The value of the shift δ hp was chosen in such a way that the 7. Transformation back to the Q-domain
probability functions are roughly zero within the range of the To express Eq.57 into the more familiar Q-domain parameters
bell curves. Note that y + δ lp ≈ y + 2.7σ y is more or less the and thus the Pc parameters, the following substitutions are
used (see also Eq.34):
maximum value related to the bell curve centered around y
as can be seen in Fig.5. u o = − g ( y o − y − δ hp ) = Qo − Q d − ∆ hp
(58)
Inclusion of this low pressure term into Eq.54 gives the u = − g ( y − y − δ hp ) = Qd − Qd − ∆ hp
following expression, valid for low- and medium y -values:
Plugging this into Eq.57 yields:
σy σy
y ≈ y + ( y o − y ) Λ ( y o − y ) + σ f2
y (y o − y )
σ σy σ
σy (56)
−1
g
u ≈ −1
g
u o Λ y ( −g1 u o ) + σ y2 f ( −g1 u o ) − y o Λ y ( −g1 u o − δ xx )
− y o ⋅ Λ ( y o − y − δ lp )
 

low P correction
which is equivalent to:

σ
u ≈ u o Λ y ( −g1 u o ) − g σ y2 f
σy
( −g1 u o ) + g4 Λ
σy
( −1
g
(u o + ∆ xx ) ) (59)

6.5 Amending the high y values where y o = 14 and ∆ xx = g δ xx = g (δ lp − δ hp ) = ∆ lp + ∆ hp .

Note that for larger values of y , i.e. outside the low-to-
medium pressure regime, Λ( y o − y ) ≈ 0 and f ( y o − y ) ≈ 0 Here it should be noted that ∆ lp = g δ lp ≈ 2.7 σ and is in
and Eq.56 yields that y → y . From the section 6.2 we know general significantly larger than ∆ hp (see Eq.47 and Fig.7).
that due to the curvature of the β(y) function, there is a small
Therefore it is safe to assume that ∆ xx = ∆ lp + ∆ hp ≈ ∆ lp , i.e.
deviation, i.e. y ≈ y + δ hp (Eq.42). This δ hp shift is caused by
the high-pressure shift has negligible effect on the low-
the quadratic components in the β-function and is a consistent
pressure correction (as expected) and Eq.59 can thus be
shift to the left away from y . Fortunately it is possible to
written as:
amend Eq.56 so that it also caters for the high pressure regime.
This can be done in several ways, but the most elegant one is
to replace y in Eq.56 with its shifted equivalent y + δ hp ,
σ
u ≈ u o Λ y ( −g1 u o ) − g σ y2 f
σy
( −g1 u o ) + g4 Λ
σy
( −1
g
(u o + ∆ lp ) )
where δ hp is now defined by − ∆ hp g (see Eq.47), i.e. a σ σy
Using that Λ y ( g1 x ) = Λσ ( x ) and g σ y2 f ( g1 x ) = σ 2 f σ (x )
function tapered in the low y value regions as is reflected by
the red curve in Fig.6. The result is: one gets:

u ≈ u o Λσ (−u o ) − σ 2 f σ (−u o ) + g4 Λσ ⎡⎣ −(u o + ∆ lp ) ⎤⎦

Moreover, Λσ ( − x ) = 1 − Λσ ( x ) and f σ (− x ) = f σ ( x ) and

*
Note that y + δ lp ≈ y + 2.7σ y is more or less the maximum value
†
related to the bell curves displayed in Fig.5. Disregarding the δ xx shift in the low-pressure term.
SPE 105139 11

thus: which varies with the applied capillary pressure. The obtained
expression is:
u ≈ u o ⎡⎣1 − Λσ (u o ) ⎤⎦ − σ 2 f σ (u o ) + g4 ⎡⎣1 − Λσ (u o + ∆ lp ) ⎤⎦ (60)
⎛ −g ⎞
Bvup (Q ) = φ exp ⎜ ⎟ (65)
By introducing here the complementary probability function ⎜ Q + ∆ − Q ⎟
⎝ g d ⎠
Ψσ , which is defined as (see also Eq.24)

∞ ∞ −x 2
where: Qd = u + Q d + ∆ hp ;
Ψ (u ) = ∫ f (x ) dx =
σ σ 1
∫e 2σ 2 σ
dx = 1 − Λ (u ) (61)
σ 2π ↑ u ≈ u o Ψ σ (u o ) − σ 2 f σ (u o ) + g4 Ψσ (u o + ∆ lp )
u u

Eq.60 can be expressed in a slightly abbreviated form:
u ≈ u o Ψ σ (u o ) − σ 2 f σ (u o ) + g4 Ψ σ (u o + ∆ lp )
This equation is de facto the solution to the upscaling problem
of the Thomeer formalism. The upscaled log-minimum entry
pressure Qd results directly from this equation (Eq.58), i.e.
Qd = u + Q d + ∆ hp
And thus also Bvup , i.e. Bvup (Q ) = φ exp
Note that in the upscaled world the log minimum entry
pressure Qd is not a single number anymore, but varies with
and ↑ u o = Q − Q d + ∆ g − ∆ hp − g4
φ - the average porosity of the rock-element
Q - the log applied pressure = ln(Pc )
(62)
Qd - the average log minum entry pressure
σ - standard deviation of the Q d 's
g - the average Pore Geometrical Factor (=2.3G )
σ g - standard deviation of the g's
σ g2
(63) ∆g = 2g - shift caused by the g-upscaling:Eq.28
∆ hp - High pressure shift (see Eq.47)
( −g
Q +∆ g −Qd ) Ψ σ - complementary probability function
f σ - normal distribution function
4

applied pressure, which enters Eq.62 through the input

parameter u o , i.e.
Qd
u o = Q − Q d + ∆ g − ∆ hp − g4 (64) 3
distribution actual

This expression follows from that fact that u o = Qo − Q d − ∆ hp

2
(Eq.58) and that Qo = Q + ∆ g − gy o (Eq.34), while y o = 14 .
∆hp
calculated

8. Summary of final results

The upscaling operation of the Thomeer parameters and
function can be summarized as follows: 0 0
0 1 2 3 4 5 6 7 8 9

(φ , g ,Qd ) ⎯⎯⎯⎯
upscaling
operation
→ (φ , g ,Qd − ∆g ) * Q =ln(Pc )

or Fig.8: The upscaled version of the log minimum entry

−g −g pressure Qd , both the actual values and the calculated ones.
Q − Qd Q +∆ g − Q d
Bv (Q) = φ e upscaling
⎯⎯⎯⎯
operation
→ Bvup (Q) = φ e The match is good. Also shown is the distribution function of
the Qd’s.

Where it should be noted that the shift ∆g is caused by the The above equations hold for a single rock system having a
upscaling of the pore geometrical factor, i.e. the g parameter. mono-modal pore-system. Moreover, correlations between
φ , g and Qd parameters are considered to be weak, which is
The up-scaled Bv can be expressed in the classical Thomeer
mainly valid for carbonates.
form, governed by the average values of the porosity and the
pore-geometrical factor (G). The log minimum entry pressure
is replaced by an effective minimum entry pressure (Q̃d), The upscaled (log) minimum entry pressure Qd is shown in
Fig.8, where the actual values are compared to the calculated
ones using Eqs.62 and 63. As can be seen the match is
*
Better is: ( B v,∞
, g , Qd ) ⎯⎯⎯
upscaling

operation
→ ( Bv , ∞ , g , Q d − ∆ g ) since in general excellent. Within the range of the Qd distribution, i.e. low to
Bv ,∞ ≠ φ , certainly not when the pore system is multimodal. medium pressures, the upscaled (log) minimum entry pressure
12 SPE 105139

Qd builds up almost linearly until it reaches a level of just

1 .0 0 0
above the average value Qd . For high pressures, i.e.
Q > Q d + 3σ (about Q>4 in Fig.8), → u o  0 and
Ψ σ (u o ) ≈ f σ (u o ) ≈ Ψ σ (u o − δ lp ) ≈ 0 , i.e. u ≈ 0 and thus

F r a c t .B u lk V o l. ( B v )
Qd ≈ Q d + ∆ hp , which slowly converges to Qd and for these

0 .1 0 0
higher pressures Bvup can be expressed as:
Bvup (Q ) ≈ φ exp ( −g
Q + ∆ g −Qd +∆ hp )

0 .0 1 0
1.0
Fract.Bulk Vol. (Bv )

0 .0 0 1
Qd

Pd
1

10

100

1000
( p s i)

10000
distribution

Pc
m ed
Fig.10: Comparison of the upscaled and average Fractional Bulk Volume
Occupied (Bv) displayed in the more conventional way. In the low pressure
regime, i.e. smaller than average Pd, the upscaled curve sets in at the lowest
minimum entry pressure and increases almost linearly.
0.0 0.0
0 1 2 3 4 5 6 7 8 9
Q =ln(Pc )
up
9. Comparison with Modeling
Fig.9: Comparing the upscaled fractional bulk volume B v with the one non To assess the accuracy of the derived approximations (Eq.65
upscaled on based on the average Thomeer parameters. Please note that the and 64) a rock element equivalent to 10,000 plugs was
average porosity is normalized, i.e. φ = 1 . The Qd distribution function is simulated.
shown as well.
FIG.10: Comparing actual Bv-upscale and calculated Bv_upscale.
The resulting upscaled fractional bulk volume Bvup of Eq.65 is The respective Qd_Tilde functions are displayed as well. Number
of plugs in rock element =10,000.
shown in Fig.9. It is compared to the normal Bv for the 14
<g>=1.4; stdev_g=0.5
4

average Thomeer parameters φ , g and Q d , i.e. 12 <Qd>=2.2, stdev_Qd=0.4

<POR>=15%
4

3
10 |Corr.coeff's |<0.4

( )
Bv-upscale

Qd_Tilde
Bvφ , g ,Qd (Q ) = φ ⋅ exp −g
Q −Q
(66) 8
2
6
Bup simmulated 2

The upscaled bulk volume shows that liquid intrusion happens 4 Bup (via Eq.65)
Qd_tilde (via Eq.64)
1

much before the non upscaled one, where intrusion occurs 2 Qd_tilde ( from Bup simm.) 1

above Qd . For Q’s beyond the highest minimum entry 0

0 1 2 3 4 5 6 7 8 9 10
0

pressures, i.e. Q > Q d + 3σ , the two curves overlap. This
reflects the picture of the Q curve in Fig.8. It also indicates
that for the Bv the small, g-upscaling induced, shift ∆g has little
effect.
Fig.10 is the more conventional display of the Bv curves of
Fig.9, i.e. Bv versus Pc in a log-log display. The low pressure
behavior of the upscaled function Bvup is highlighted in this
figure. The liquid intrusion for the large rock volumes
happens at a much lower pressure than what can be expected
from the average rock properties only. This means that in an
oil reservoir there will be more oil present at the lower heights
above the free water level. Than from single core plug data
could be expected and modeled. For greater heights the
upscaled and non upscaled curves more or less overlap.
Q=ln(Pc)
d The following parameters were used:
φ = 15%;
G = 0.6 → g = 2.3 ⋅G = 1.38; σ G = 0.2 → σ g = 0.46
Q = 2.2; σ Qd ≡ σ = 0.4
Number of "plugs" = 10,000
Corr.Coeff (φ ,g,Qd ) < 0.4 ( ≈ negligible)
plug − plug correlation coeff. < 0.9 and has no effect.
The results are displayed in Fig.10. The difference between
Bvup,actual and Bvup,calculated is small.
SPE 105139 13

Conclusion Λσx (x ) Probability function (mean x ; stand.dev. σ )

Here a direct and closed form approximation to the upscaling
Ψσx (x ) Complementary probability function
problem of mercury injection capillary pressure data to
arbitrary large reservoir elements is presented. The method is β (y ) intermediate function = exp( y1 ) . See Appendix.
based on standard Thomeer functions and a statistical σ g2
consistent formalism has been derived based on Gaussian ∆g ≈ 2g shift in Q-domain caused by variation in g (G).
distributions of the key parameters. ∆ hp high pressure shift in Q-domain
The result retains the familiar form of the Thomeer
function while replacing single plug values with statistical y = g1 (Q + ∆ g − Q d )
- coordinate
properties of these values within a petrophysical rock type. y = g1 (Q + ∆ g − Q d ) transformation from
These results will now allow improved integration of core
y o = g1 (Q + ∆ g − Q d ,o ) Q- to y-domain
plug and log data processes for saturation-height function
calibration to the remaining variables, i.e. subsurface y = g1 (Q + ∆ g − Q d' )
wettability, reservoir fluid properties and electrical log
interpretation parameters. u o = Qo − Q d − ∆ hp intermediate coordinate transformation
It is anticipated that statistically robust sample sets of u = Qd − Q d − ∆ hp intermediate coordinate transformation
MICP data within petrophysical rock types will become much
σ g2
more common for these purposes and progress will be δy = 2g 2
- Shift in y -domain
accelerated in understanding subsurface fluid distributions,
free-water level determinations and geometries, and that δ hp - High pressure shift in y -domain
wettability references will improve reservoir analysis. σ y = σg - Standard deviation in y -domain

Nomenclature
MICP Mercury Injection Capillary Pressure References
Pc Capillary pressure (applied Hg pressure) in psi 1. Thomeer, J.H.M., 1960, Introduction of a Pore
Pd Minimum entry pressure in psi Geometrical Factor Defined by a Capillary Pressure
Curve, Petroleum Transactions, AIME, Vol 219, T.N.
Bv , ∞ Fract. Bulk Vol. occupied by Hg at Pc=∞
2057, 354-358 .
Bv Fract. Bulk Vol. occupied by Hg 2. Ekrann, S., 1999, Water Saturation Modeling: An
Bv ,i Fract. Bulk Vol. occ. for core plug i Upscaling Point of View, SPE 56559.
3. Clerke, E.A., 2003, Beyond Porosity-Permeability
φ , g ,Qd
Bv Fract. Bulk Vol. occ. For plug (φ , g ,Q d ) Relationships – Determining Pore Network Parameters
B up
Upscaled Fract. Bulk Vol. occ. by Hg for the Ghawar Arab D using the Thomeer Method,
v
GeoFrontier, Vol 1, Issue 3, September.
b up
g Normalized F.B.V. occ. by Hg Bvup φ 4. Clerke, E.A., Private communications, 2005
E[x] Expectation value of x=<x>
G Pore Geometrical Factor (PGF)
g PGF in Q-domain of a core plug (=2.3G)
g Average of all g’s in the large rock
Acknowledgements
σg Stand. deviation of the g’s. The author would like to acknowledge helpful discussions with Ed
σ Stand. deviation of the Qd’s. Clerke and Craig Phillips and he wishes to thank Saudi Aramco for
Q Applied Hg pressure in the Q-domain: Q = ln(Pc ) allowing him to publish this work.
Qd Minimum entry (Hg) pressure in the Q-domain
Qd Average minimum entry pressure in the Q-domain
Qd Upscaled equivalent of Qd (not a constant)

Qd Preliminary notation (= Qd − ∆ g )
∆o Shift in Q-domain caused by Q-integral
Qd ,o Shifted min. entry pressure (= Q − ∆o + ∆ g )
Ω normalized distrib.func. in (φ , g ,Q d ) space
ψ .. normalized 1-D distribution functions
φ porosity of a core plug
φ average porosity of all plugs in a large rock element
σ
f x (x ) Normal distrib. (mean x ; stand.dev. σ )
14 SPE 105139

Appendix : Properties of the function β(y) Extremes:

The exponential part of the Thomeer function in the Q-domain First derivative
(see Eq.3) can be simplified by the β-function as introduced in β ' ( y = y a ) ⇒ β '' ( y a ) = 0 ⇒ y a = 1
2
section 6.1:
Point ya corresponds also with the inflection point on
⎪⎧e y for y > 0
−1

β (y ) = ⎨ the flank of the β(y) function. Note that β ' ( 12 ) ≈ 0.54

⎪⎩ 0 for y ≤ 0
The function is shown in the figure below and converges Second derivative
slowly to 1 for y→∞. 1 6
β '' ( y = y b ) ⇒ β ''' ( y b ) = 0 ⇒ − +6 = 0
y b2 y b
1 4

⎧⎪ y b+ = 3+1 3 ≈ 0.211
β (y) Which has two solutions: ⎨ −
⎪⎩ y b = 3− 3 ≈ 0.789
1
0.75
β ''(y) 3

y b+ corresponds to the point where the β(y) function

β (y) and β '(y)

0.5 2
starts to grow, i.e. β (0.211) ≈ 0.009 and
β ''(y)
β ( y < y b+ ) ≈ 0 .
0.25 β '(y) 1

inflection
point
y b-
0 0 For low y’s β(y) can be approximated by a tangent line
0 y b+ 0.5 1 1.5 2 2.5 3
through the inflection point, i.e.
ya y
-0.25 -1
β ' ( 12 ) ( y − y o )

Derivatives: This line crosses the y-axis at yo=¼, see figure.

∂β 1
β (y ) =
'
= β (y ) 2 0.75
∂y y
β ' (½ )·(y-¼ )
∂ β
2
1 ⎛1 ⎞
β '' ( y ) = 2
= β (y ) 3 ⎜ − 2⎟ 0.5

∂y y ⎝y ⎠ β '(y)
β (y)
β(y) and β'(y)

∂3 β 1 ⎛ 1 6 ⎞
β ''' ( y ) = 3
= β (y ) 4 ⎜ 2 − + 6⎟ 0.25

∂y y ⎝y y ⎠
yb
+ inflection point

0
0 0.25 0.5 0.75 1 1.25 1.5

yo=¼
y
-0.25