Title:

The effects of confining pressure, porefluid salinity and saturation on the acoustic properties of sandstones.

Modern seismic data acquisition and processing methods now enable scientists to
extract information on both the stratigraphy and the physical properties of subsurface
rocks. Laboratory acoustic measurementsa llow the physical conditions to be precisely
measured and controlled. In the present study, P and Swave velocities (Vs, VS) and
attenuations (1000/Qp, 1000/Qs) were measured in a range of sandstones using the
ultrasonic pulseecho technique, at effective pressures of 5 MPa to 60 MPa. The
measurement accuracy is ±0.3 % for velocity and ±0.1 dB/cm for attenuation using this
method.
Velocities and quality factors( Q) fall with decreasinge ffectivep ressure,a nd the
relationships are described by the empirical equationsV =A+KPB C71' and
Q=AB e7DP , where P is the effective pressure and A, K, B, and D are the regression
coefficients (D=0.115±0.016 and 0.048±0.010 for V and Q, respectively). Velocity
and Q can therefore be extrapolated to pressures beyond the experimental range. The
Biot, Gassmann, and unrelaxed porefluid models of seismic wave propagation in porous
media fail to explain the pressuredependenceo f the velocities.
The difference between the experimental and Biot model predictions of the rate of
change in Pwave velocity with pore fluid salinity (dVýdM) increases with percentage
clay content (C) of the rock at the approximately linear rate of 0.95 m/s/mol. There is
no clear relationship for dVs/dM. In clean sandstones there is a close agreement between
the experimental results and Biot model predictions for dVP/dM, but the agreement breaks
down when C>5%. This suggests that changes in the porefluid salinity alter the frame
bulk and shear moduli of sandstones. Attenuation is generally independent of porefluid
salinity.
Attenuation and velocity are often strongly dependent on the degree of porefluid
saturation. A study of nine samples shows that 1000/Qp exhibits a resonance peak at midrange
saturations (SW av 30 % to SW = 70 %) in most samples, and 1000/Qs shows similar
behaviour in several of these. For porosities greater than 13%, the normalised amplitudes
of the peaks in Pwave and bulk attenuation are correlated to porosity; the latter increases
at a linear rate of 0.98 per percentage increase in porosity. These data suggest that
attenuation reaches a maximum when the gas/water mixture is neither too compressible nor too incompressible. The Biot/squirt (BISQ) theory inadequately models the saturation
dependence of 1000/Qp and Vp in a sample at low confining pressure. Vp falls with
decreasing saturation between SW =100 % to SW  50 %; below SW = 50 %, the behaviour of
Vp is dependent on the confining pressure. Vs generally increases with decreasing
saturation over the entire saturation range in all samples. The unrelaxed porefluid model
of Mavko and NolenHoeksema (1994) describes the Vp data reasonably well in most
samples using low wetting fractions (< 15 %), which indicates that the pore fluid is
unrelaxed at both the grain and sample scales. The wetting pore fluid becomes unrelaxed
at high frequencies and/or low permeabilities. The V. data are poorly described by the
model, possibly due to matrix softening by the wetting fluid.
The experimental data have indicated significant shortcomings in the mathematical
models of seismic wave propagation in reservoir rocks. The data highlight important
aspects of wave propagation that must be addressed in revised theories.
