Modeling and inversion of stress-induced multicomponent seismic time shifts

Abstract

Subsurface pressure and temperature variations can alter rock properties both near and relatively far from the disturbance, causing detectable changes in seismic traveltimes. In this thesis, I first use traveltime variations to study velocity changes around a heated prototype nuclear waste storage tunnel. Then I model and invert compaction-induced multicomponent time shifts from depressurizing petroleum reservoirs. Heaters inside the tunnel replicate the thermal output of decaying radioactive waste, heating the tunnel over two years and maintaining a constant temperature for another two years. Time-lapse velocity models were constructed using temperature-dependent velocity data for granite and thermal profiles from boreholes in the tunnel wall. Matching check-shot and modeled waveforms indicate that the tunnel temperature can be monitored using seismic data. Further, the smooth, unperturbed velocity field lacks spatial perturbations, suggesting that no fluid or steam was present around the tunnel near the receiver array during the experiment. However, the combination of changing velocities and non-elastic, stress-induced acoustic emissions near the tunnel crown suggest that damage to the rock may occur. To study time shifts around a compacting reservoir, I employ geomechanical modeling of the compaction-induced stress/strain fields. Strain-dependent stiffness perturbations are obtained from the nonlinear theory of elasticity. Then full-waveform multicomponent seismic data are generated by finite-differences and used to estimate the time shifts of P-, S-, and PS-waves. P-wave time shifts are strongly influenced by compaction-induced velocity anisotropy around the reservoir. S-wave anisotropy is almost negligible, but S-wave shifts are 2-3 times larger than those of P-waves. PS-wave time-shift behavior significantly varies with the reflection point. Spatial time-shift distributions are exploited to study the sensitivity of each wave type to reservoir pressure ([Delta]P) and length (L). The analysis shows that time shifts are nonlinear in stiffness and, therefore, reservoir pressure for large values of [Delta]P. A hybrid global/gradient inversion algorithm requiring neither Jacobian or Hessian computations is developed to invert noise-contaminated time shifts for reservoir pressure and length. Numerical testing shows that P-wave shifts from the top of the reservoir produce accurate estimates of [Delta]P and L, even for noisy data. If multicomponent data are available, S-wave time shifts and those of all wave types combined from reflectors beneath the reservoir yield accurate estimates of [Delta]P. The strain fields for the best inverted models do not deviate from the actual strains by more than 20%. The thesis results suggest which wave types and reflectors provide the most accurate estimates of changing reservoir parameters and compaction-induced stress/strain fields. This can provide important information for making drilling decisions and improving hydrocarbon recovery.