Finite element modeling of seismic attenuation due to fluid flow in partially saturated rocks
Beatriz Quintal*, ETH Zurich, Holger Steeb, Ruhr-University Bochum, Marcel Frehner, University of Vienna,
and Stefan M. Schmalholz, ETH Zurich
Summary
The finite element method is used to solve Biot’s quasi-
static equations of consolidation. We perform 1D and 2D
numerical creep tests of partially saturated porous rocks to
calculate the frequency-dependent seismic attenuation and
phase velocity from the modeled stress-strain relations. The
resulting attenuation and velocity dispersion are due to
fluid flow induced by pressure differences between
mesoscopic-scale regions of the rock fully saturated with
different fluids (White’s model). Comparisons of our
numerical results with analytical solutions show accuracy
for a wide range of frequencies. The algorithm is applied to
a 1D partially saturated rock with a random distribution of
saturation. We show that the numerical results for the
random distribution can be approximated with a volume
average of analytical solutions for periodic media.
Introduction
Attenuation of seismic waves in partially saturated rocks is
of great interest because it has been observed that gas and
oil reservoirs often exhibit high attenuation (e.g., Dasgupta
and Clark, 1998; Rapoport et al., 2004), especially at low
frequencies (Chapman et al., 2006). Data, from both
laboratory and field, and theoretical work show that
attenuation can be related to an increase in reflectivity in
the low-frequency range (Korneev et al., 2004; Quintal et
al., 2009). Goloshubin et al. (2006) showed three examples
of field data in which oil-rich reservoirs exhibit increased
reflectivity at low seismic frequencies (around 10 Hz).
At low seismic frequencies, wave-induced fluid flow on the
mesoscopic scale is presumably the major cause of wave
attenuation and velocity dispersion in partially saturated
porous rocks (e.g., Norris, 1993; Johnson, 2001; Pride and
Berryman, 2003a, b). The mesoscopic scale is the scale
much larger than the pore size, but much smaller than the
wavelength. White (1975) and White et al. (1975) were the
first to introduce the wave-induced fluid flow mechanism
for a 3D model of a water-saturated medium with spherical
gas-saturated inclusions and a 1D layered model. In
White’s model, a partially saturated rock is represented by
a poroelastic solid with regions fully saturated by one fluid
and regions fully saturated by another fluid. Wave-induced
fluid flow is caused by pore pressure differences between
the two regions. Dutta and Odé (1979a, b) showed that
wave-induced fluid flow can be modeled using Biot’s
equations (Biot, 1962) for wave propagation in poroelastic
media with spatially varying petrophysical parameters.
Several theoretical studies, based on White’s model and
Biot’s theory (Biot, 1962), provide various closed-form
analytical solutions for seismic attenuation in porous
saturated media with periodic mesoscopic-scale
heterogeneities of particular geometries, such as layered
media or media with spherical inclusions (e.g., Johnson,
2001; Pride and Berryman, 2003a, b). There are also
closed-form analytical solutions for randomly layered
media (e.g., Gurevich and Lopatnikov, 1995), however,
they are restricted to infinite media and to particular
autocorrelation functions. Müller and Gurevich (2005)
showed that significant differences in the magnitude and
frequency dependence of attenuation are caused by only the
use of different autocorrelation functions. Additionally, in
the low-frequency limit, 1/Q (Q is the quality factor, 1/Q is
a measure of attenuation) scales differently in infinite
randomly layered media, compared to periodically layered
media or finite randomly layered media. For infinite
random media, 1/Q is proportional to the square root of
frequency, while for periodic and finite random media it is
proportional to frequency (Müller and Rothert, 2006).
Thus, stable and accurate numerical solutions for seismic
attenuation in porous saturated media with mesoscopic-
scale heterogeneities are required, for example, for: (i)
heterogeneities with complicated geometries, (ii) finite
random media with arbitrary distribution pattern, or (iii)
media containing more than two heterogeneities, such as
partial saturations with more than two fluids.
Calculating seismic attenuation due to wave-induced fluid
flow with numerical algorithms for wave propagation in
poroelastic media (e.g., Zhu and McMechan, 1991) is
computationally inefficient because wave propagation,
fluid flow and fluid pressure diffusion occur on different
time scales. An efficient method is a quasi-static creep test,
suggested by Masson and Pride (2007), in which they
solved Biot’s equations (Biot, 1962) for wave propagation
in poroelastic media with the finite difference method. In
this study, we performed similar quasi-static creep tests for
calculating seismic attenuation as suggested by Masson and
Pride (2007), however, we solved a simpler mathematical
problem, Biot’s equations of consolidation (Biot, 1941), in
which inertia forces do not play a significant role and
therefore are excluded. Attenuation due to wave-induced
fluid flow is controlled by fluid pressure diffusion. For
calculating the amount of attenuation, it is sufficient to
model only the pressure diffusion. We used the finite
element method to solve Biot’s equations of consolidation
in the u-p formulation (Zienkiewicz and Shiomi, 1984). We
show that our numerical scheme is powerful and accurate
in calculating attenuation and velocity dispersion due to the
2564
SEG Denver 2010 Annual Meeting
© 2010 SEG