Microseismic Monitoring and Geomechanical Modelling of CO2 - bris
Microseismic Monitoring and Geomechanical Modelling of CO2 - bris
Microseismic Monitoring and Geomechanical Modelling of CO2 - bris
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
6.5. COMPARISON OF ROCK PHYSICS MODELS<br />
that the response <strong>of</strong> velocities to stress change is in fact nonlinear, so a linear interpolation is in fact<br />
a poor approach to take.<br />
The approach most commonly used in industry is the R-factor model <strong>of</strong> Hatchell <strong>and</strong> Bourne (2005).<br />
This model assumes that the fractional change in vertical P-wave velocity is proportional to the vertical<br />
strain (with the R-factor being the constant <strong>of</strong> proportionality in the equation dV P z /V P z = −Rε zz ).<br />
Vertical P-wave velocities are the most commonly measured property in conventional seismic surveys,<br />
<strong>and</strong> changes in vertical stress (<strong>and</strong> strain) will be largest geomechanical effect above a compacting<br />
reservoir. The R-factor approach was developed to address this particularly relevant subset <strong>of</strong> geomechanical<br />
scenarios. Hatchell <strong>and</strong> Bourne (2005) found that R-factors were reasonable consistent<br />
across a number <strong>of</strong> sites, although R-factors for rocks experiencing compressive strain were found to<br />
be 5 times smaller than for rocks under extensional strain. However, more recent studies have shown<br />
order-<strong>of</strong>-magnitude variations in R-factors depending on lithology (e.g., Staples et al., 2007; De Gennaro<br />
et al., 2008). More importantly, R-factors have been found to be dependent on the stress path<br />
(Holt et al., 2008), <strong>and</strong> on the magnitude <strong>of</strong> the applied stress (Pal-Bathija <strong>and</strong> Batzle, 2007).<br />
The different R-factors required for extension <strong>and</strong> compaction, for different applied stress magnitudes<br />
<strong>and</strong> for different triaxial stress states, all for the same rock, means that this approach does<br />
not lend itself to model scenarios where the stress changes during production are not known in advance.<br />
This is because the R-factor model does not adequately describe the full, triaxial, anisotropic,<br />
nonlinear response <strong>of</strong> seismic velocities - it is limited to vertical strain <strong>and</strong> vertical P-wave velocities.<br />
Therefore it is not capable <strong>of</strong> dealing with the observed phenomena that fall beyond its remit. Unfortunately,<br />
these issues are very much within my remit when dealing with the geomechanical models<br />
developed in Chapter 5, so I do not use the R-factor approach. Nevertheless, because it cuts through<br />
a complicated system to leave one <strong>of</strong> the key 4D seismic parameters (vertical timeshift) correlated via<br />
one parameter (the R-factor) to one geomechanical observable (vertical strain) it remains as attractive<br />
approach within the industrial sector.<br />
Like many rock physics models, the R-factor approach attempts to fit the observed nonlinear stressvelocity<br />
response with a linear model, which means that it is limited in its applicability. Furthermore,<br />
the model only considers the vertical P-wave velocity response to vertical strain. Observations show<br />
that vertical P-wave velocity is also modulated by horizontal deformation, although this is a second<br />
order effect (this result can be derived from equation 6.11). More importantly, even in reflection surveys,<br />
seismic waves do not travel vertically, but through a range <strong>of</strong> inclinations about the subvertical.<br />
Therefore the R-factor approach tends to break down for longer <strong>of</strong>fset arrivals (Herwanger, 2007).<br />
The third-order elasticity model developed by Prioul et al. (2004) includes the effects <strong>of</strong> triaxial<br />
stress changes on the full anisotropic stiffness tensor. This means that the effects <strong>of</strong> stress-induced<br />
anisotropy, <strong>and</strong> variations in larger <strong>of</strong>fset timeshifts, can be incorporated. This model takes a mathematical<br />
approach, including the third-order terms that are usually discarded in deriving conventional<br />
linear elastic theory. Therefore, this model does not directly address the microstructual properties <strong>of</strong><br />
the rock, although it is capable <strong>of</strong> modelling the full stiffness tensor. However, to facilitate inversion<br />
from ultrasonic data, Prioul et al. (2004) have to assume an isotropic third order tensor, which in<br />
131