first break volume 22, February 2004

technical article

Pore fluid and porosity mapping from seismic

Jack Dvorkin and Salman Alkhater*
Abstract
We use rock physics to map pore fluid and porosity from
seismic data in a vertical section between two wells. First,
well log data are used to establish an effective-medium model
that links the impedance to pore fluid and porosity. Next,
stacked seismic data are used to produce P-wave impedance
inversion. Finally, the rock physics transform is applied to
the impedance section to identify pore fluid and produce a
porosity section.

Introduction
For decades, the main use of seismic data has been to delineate sedimentary bodies and tectonic features in the subsurface. The mission of exploring inside the geological body is a
relatively recent development. Mapping porosity, lithology
and other reservoir bulk properties inside the geological body
has become possible due to the recent dramatic improvement
in seismic acquisition, imaging and inversion quality, as well
as the accompanying advances in rock physics.
Rock physics provides transforms between a reservoir's
elastic properties and its bulk properties and conditions,
including porosity, lithology, pore fluid and pore pressure.
Such transforms are known as trends. Trends are built from
controlled experiments where both the elastic and bulk prop-

erties of rock are measured on the same samples under the

same conditions. The most commonly used source of such
experimental data in modern rock physics is the borehole
measurement. For example, an empirical impedanceporosity trend developed from sonic, density and porosity curves
can be applied to a seismic acoustic impedance volume in
order to map porosity in 3D.
However, it is always advantageous not only to find an
empirical trend but also to understand the physical laws that
determine the trend or, in other words, find an appropriate
effective-medium model. Such rationalization of an empirical
trend by effective-medium modelling generalizes the trend,
determines the domains of its applicability, and thus reduces
the risk of using the trend outside the immediate data range.
Here, we illustrate the rational-rock-physics approach by
mapping porosity in a large producing gas/oil reservoir. Well
data are used to establish a transform from impedance to
porosity, based on rock-physics theory. This transform is
then applied to a vertical impedance section obtained from
stacked seismic data via inversion.
The reservoir under examination consists of relatively
soft sands. As a result, the acoustic impedance of the gas-saturated sand is much lower than that of the oil- and water-saturated sand. This large impedance difference allows us to

Figure 1 The seismic section (full stack) between two vertical wells. The reservoir is located between reflectors A and B.
Reflector C marks the transition from gas to oil.
* Stanford Rock Physics Laboratory

2004 EAGE

53

technical article

first break volume 22, February 2004

Figure 2 Well-log data for Well 1 (left) and Well 2 (right). The well-log curves are (from left to right): gamma-ray (API);
hydrocarbon saturation (fraction); measured P-wave impedance (km/s g/cc); total porosity (fraction); P-wave impedance
(km/s g/cc) calculated for 100% water saturation. The porosity and impedance frames also show the upscaled curves that
are the running mean average for porosity and the Backus average for impedance with one-quarter wavelength used as an
averaging window.
identify the pore fluid from P-wave data only, without using
offset information. As a result, we map both pore fluid and
porosity, using only stacked seismic data.

Seismic data
The seismic line under examination (Fig. 1) connects two
vertical wells and was extracted from a 3D seismic volume in
the North Sea Troll Field. The data were sampled at 2 ms
intervals and stacked at all offsets. The two wells are located
at CDP gathers 3 and 190. The two-way traveltime window
of interest is between 1400 and 1900 ms. The reservoir is
located between two seismic horizons (A and B in Fig. 1).
The gasoil contact is located at the flat reflector C. The
relief of the reservoir changes abruptly in the middle of the
section, which is typical for the step-like offshore sediment
topology in the Troll Field.

Well-log data and rock-physics diagnostics

Both the wells that bound the seismic line are producers
and they penetrate an extensive gas cap and the underlying
oil and water intervals (see well-log curves in Fig. 2).
The intervals in both wells are dominated by very highimpedance and low-porosity calcite-cemented layers. These
layers are likely to affect the seismic response of the reservoir significantly.
The pore-fluid system in the reservoir includes gas, oil
and water. We assume that there is only gas and water above
the gasoil contact and only oil and water between the

54

gasoil and oilwater contacts. In other words, gas and oil

do not coexist in the reservoir.
The in situ physical properties of the reservoir fluids, as
calculated from gas and oil gravity, gas-to-oil ratio, water
salinity, and pressure and temperature using the Batzle and
Wang (1992) relationships, are summarized in Table 1.
To develop an impedanceporosity transform, we first
bring the data to the common fluid denominator, i.e. calculate the P-wave impedance in both wells for 100% water saturation using the P-wave-only fluid-substitution method of
Mavko et al. (1995). This method replaces the traditional
Gassmanns (1951) fluid substitution where shear-wave data
are not available.
The water-substituted impedance curves are also displayed in Fig. 2. These curves practically mirror-image the
total porosity curves in both wells, which implies the existence of an impedanceporosity trend. The impedanceporosity crossplots show that these trends do exist (Fig. 3). They
are especially clear in the gas-cap intervals of both wells.

Table 1 Physical properties of gas, oil and water used in the

study

2004 EAGE

technical article

first break volume 22, February 2004

In order to describe the impedanceporosity trends

shown in Fig. 3, we use the uncemented sand model of
Dvorkin & Nur (1996) and the constant cement model of
Avseth et al. (2000). The red curves in the impedanceporosity crossplots in Fig. 3 are from the first model for sand with
clay content of 0%, 10% and 20%. The black curves are
from the constant-cement model.
By drawing the effective-medium curves, we are able to
identify two principal branches in the data. The first branch
is comprised of high-porosity relatively stiff sands that have
very small amounts of contact quartz cement (the black
model line in Fig. 3). The second branch is comprised of
uncemented sands with very small amounts of clay. This
trend separation is also common in the Troll Field and is
determined by grain sorting. Large well-sorted grains may
have small quantities of cement covering them while the
presence of even a small amount of clay inhibits cement formation (Avseth et al. 2000).
We conclude that the impedanceporosity behaviour in
the reservoir under examination can be accurately described
by both the uncemented and the slightly cemented sand models, depending on the type of sand.

Upscaling
The seismic wavelength is usually much larger than the scale
of rock-property variation observed in well-log data. Let us
assume that the average P-wave velocity is 2.5 km/s and the
dominant frequency is 50 Hz. Then the wavelength is 50 m
and the quarter-wavelength is 12.5 m. In order to understand
how the rock properties observed from seismic are related to
those derived from well logs and also how the rock-physics
transforms relevant to well-log data hold at the seismic scale,
we conduct upscaling using the quarter-wavelength as an
averaging window.
The two rock properties to be linked to each other are (a)
the total porosity and (b) the P-wave impedance. Therefore,
for both wells, we (a) calculate the average porosity by applying the running mean filter to the data and (b) calculate the
effective impedance by using the Backus (1962) average. These

mean porosity and effective impedance curves are displayed in

Fig. 2, together with the log-scale measurements. The effective
(fluid-substituted) impedance is crossplotted versus the average porosity in Fig. 3, next to the log-scale crossplots.
The upscaled crossplots appear to be even tighter than
the log-scale crossplots, especially in the gas zone of Well 1.
After upscaling, the data points move from the constant
cement line on to the clean uncemented sand line, most probably due to the effects of the Backus averaging on the impedance (the Backus average is the geometric average of the
compressional moduli and as such gives more weight to the
softest layers in the sequence (G. Mavko, pers. comm.)).
The upscaled data lie between the 0% clay and 20% clay
uncemented sand lines. This spread determines the accuracy
of porosity determination from seismic (about 0.03 porosity units).

Identifying pore fluid

In the soft sand under examination, the effect of pore fluid
on the P-wave impedance is so large that it allows us to discriminate the gas-saturated sands from the liquid-saturated
sands by using simple impedance cut-off. Consider Fig. 4
(left) where the upscaled impedance is plotted versus the
upscaled total porosity for both wells and colour-coded
according to the pore fluid.
The impedance in the gas cap (red symbols) predominantly lies below 6 km/s g/cc while the impedance of the oiland water-saturated intervals lies above the 6 km/s g/cc
mark. This clear cut-off allows us to identify the pore fluid
from P-wave data by following a simple rule, i.e. gas is present where the P-wave impedance is lower than 6 km/s g/cc
and oil or water is present where it is higher than 6 km/s g/cc.
Note that this cut-off is not universal. It is valid for the soft
sand under examination and is field and facies specific.

Transform from impedance to porosity

In order to derive porosity from impedance, we use the transform given by the uncemented sand model for 10% clay content that describes, on average, the trends present in the data

Figure 3 Impedanceporosity crossplots of water-substituted well-log data at the original log scale as well as upscaled. The
curves in the crossplots are from effective-medium models. The red curves are from the uncemented sand model with clay
content of 0%, 10% and 20%. The black curve is from the constant cement model. In the models, the critical porosity is
0.4, the coordination number is 8 and the cemented porosity is 0.375.

2004 EAGE

55

technical article

first break volume 22, February 2004

Figure 4 Left: Upscaled impedance versus upscaled total

porosity in both wells. The red symbols come from the gas
cap while the blue symbols are from the oil/water zone. The
horizontal bar is the cut-off used for pore-fluid identification.
The curves are given by the uncemented sand model for 10%
clay content, for water (top) and gas (bottom). Right:
Impedance to porosity transform as given by equations (1)
and (2) and the cut-off between the gas- and liquid-saturated
intervals.

Figure 6 Porosity from seismic (blue) and upscaled porosity

from well logs (red) versus two-way traveltime for Well 1
(left) and Well 2 (right). The horizontal bars mark the gasoil
and oilwater contacts.
(Fig. 4, left). The exact model equations are algebraically
involved and therefore it is inconvenient to apply these equations to the seismic impedance inversion data directly. We
simplify these transforms by least-squares polynomial fitting.
The resulting equations are

= 0.744 0.515IP + 0.009IP2

(1)

for IP < 6 (gas cap), and

= 1.065 0.185IP + 0.009IP2

(2)

for IP 6 (oil/water zone).

These equations, combined with the 6 km/s g/cc cut-off, are
plotted in Fig. 4 (right).

Impedance inversion and porosity mapping

Figure 5 Impedance inversion section (top) and predicted

porosity section (bottom).The inserted strips in the upper
frame mark the impedence in the wells

56

To obtain a porosity map using equations (1) and (2), we first

conduct impedance inversion on the fully stacked seismic
data using the HampsonRussell STRATA inversion package. The impedance inversion section is shown in Fig. 5
(top). As expected, the seismic impedance closely matches
the impedance in the wells. In this section, we can clearly
identify the two horizons that bound the reservoir as well as

2004 EAGE

technical article

first break volume 22, February 2004

the flat reflector that marks the gasoil contact. The impedance dramatically increases across this contact.
Next we apply the impedanceporosity transform functions, given by equations (1) and (2), to the impedance section. The resulting total porosity section is shown in Fig. 5
(bottom). The predicted porosity curves at the wells are compared to the upscaled well-log data in Fig. 6. The seismic
porosity curves closely track the upscaled well-log porosity
curves. The mismatch present in some sections of Well 1
is probably due to the scatter of data around the impedanceporosity transform curves (Fig. 4).
Note that the porosity section is qualitatively different
from the impedance section, i.e. it does not reflect the sharp
horizontal impedance contrast at the flat gasoil contact
reflector (about 1680 ms two-way traveltime). This is
because before using the impedanceporosity transforms, we
identified the pore fluid and then applied the transforms
selectively, according to equations (1) and (2).

Backus, G.E. [1962] Long-wave elastic anisotropy produced

by horizontal layering. Journal of Geophysical Research 67,
44274440.
Batzle, M. and Wang, Z. [1992] Seismic properties of pore
fluids. Geophysics 57, 13961408.
Dvorkin, J. and Nur, A. [1996] Elasticity of high-porosity
sandstones: Theory for two North Sea datasets. Geophysics
61, 13631370.
Gassmann, F. [1951] ber die Elastizitt porser medien.
Vier. Der Natur. Gesellschaft in Zrich 96, 123.
Mavko, G., Chan, C. and Mukerji, T. [1995] Fluid substitution: Estimating changes in Vp without knowing Vs.
Geophysics 60, 17501755.

ADVERTISING

Conclusion
Model-based rock physics analysis of well-log data combined with seismic impedance inversion allows us to see
inside a reservoir and, by producing reservoir property volumes, to go beyond outlining geobodies and mapping lithofacies. In the framework of this approach, it is paramount to
understand the factors that affect the trends present in the
data and to describe these trends by an effective-medium
model rather than by a direct statistical fit.
We have shown that, in soft sand, it is possible to discriminate gas-saturated intervals from liquid-saturated intervals by using only P-wave data. This result, however, does
not eliminate the need for AVO analysis as the primary tool
for hydrocarbon detection. The reason is that in certain environments, water-saturated shale may be as soft as gas-saturated sand and the latter can only be delineated on the basis
of a low Poissons ratio.
The resolution of reservoir properties produced by integrating rock physics and seismic still does not exceed seismic
resolution. As a result, it should not be expected that, without additional stringent assumptions, reservoir properties
can be mapped at sub-seismic resolution. By using the deterministic approach described here, we only can image average
properties, such as average porosity.

FIRST BREAK
NEAR SURFACE GEOPHYSICS
PETROLEUM GEOSCIENCE
GEOPHYSICAL PROSPECTING

Laura Piontek
Tel.: +31 30 635 40 55
Fax: +31 30 634 35 24
E-mail: lp@eage.nl

Acknowledgements
The data and technical advice, as well as part of the funding
for this study, were supplied by Norsk Hydro. This study
was also supported by the Stanford Rock Physics Laboratory
and Saudi Aramco.

References
Avseth, P., Dvorkin, J., Mavko, G. and Rykkje, J. [2000]
Rock physics diagnostic of North Sea sands: Link between
microstructure and seismic properties. Geophysical Research
Letters 27, 27612764.

2004 EAGE

For more information:

WWW.EAGE.NL
57