We analyze the stability and accuracy of the coarse-grain memory variable technique used for viscoelastic wave-field simulations. Our analysis shows that the general behavior of the coarse-grain system is well described by effective parameters ( M E and Q E ) that are derived from the harmonic average of the moduli over the volume of the coarse-grain cell. The use of these effective parameters proves essential for analyzing the performance and accuracy of the coarse-grain system for Q values less than about 20. A necessary stability condition for the coarse-grain system requires that the weighting coefficients be bounded between 0 and 1. Specifying the weights using the approach of Day and Bradley (2001) satisfies this condition for Q values of about 3 and larger; however, using unconstrained optimization techniques will often produce weights that violate this condition at much higher Q values. We also derive a variation of the original coarse-grain methodology called the element-specific modulus (ESM) formulation in which each element of the coarse-grain cell uses a different unrelaxed modulus. We demonstrate that the accuracy of the coarse-grain system for Q values lower than about 20 is significantly improved with the ESM formulation without increasing the computational cost. Finally, we present a technique for optimizing the accuracy of the coarse-grain system for very low Q based on the use of the effective quality factor ( Q E ). We demonstrate that using conventional optimization techniques that do not employ the effective parameter Q E will actually degrade the accuracy of the coarse-grain system.
Manuscript received 21 March 2002.
[1]
Walter H. F. Smith,et al.
Free software helps map and display data
,
1991
.
[2]
Peter Moczo,et al.
Testing four elastic finite-difference schemes for behavior at discontinuities
,
1993
.
[3]
S. Orszag,et al.
Approximation of radiation boundary conditions
,
1981
.
[4]
Robert W. Graves,et al.
Simulating seismic wave propagation in 3D elastic media using staggered-grid finite differences
,
1996,
Bulletin of the Seismological Society of America.
[5]
Einar Kjartansson,et al.
Constant Q-wave propagation and attenuation
,
1979
.
[6]
Steven M. Day,et al.
Efficient simulation of constant Q using coarse-grained memory variables
,
1998,
Bulletin of the Seismological Society of America.
[7]
Efficient 3-D viscoelastic modeling with application to near‐surface land seismic data
,
1998
.
[8]
J. Bernard Minster,et al.
Numerical simulation of attenuated wavefields using a Padé approximant method
,
1984
.
[9]
Don L. Anderson,et al.
Velocity dispersion due to anelasticity; implications for seismology and mantle composition
,
1976
.
[10]
Martin J. Siegert,et al.
EOS Trans. AGU
,
2003
.
[11]
M. Carcione,et al.
Wave propagation simulation in a linear viscoacoustic medium
,
1997
.
[12]
Joakim O. Blanch,et al.
Viscoelastic finite-difference modeling
,
1994
.
[13]
Christopher R. Bradley,et al.
Memory-Efficient Simulation of Anelastic Wave Propagation
,
2001
.
[14]
M. Korn,et al.
Incorporation of attenuation into time-domain computations of seismic wave fields
,
1987
.