Time-lapse seismic attenuation as a tool for monitoring hydrocarbons and CO2 in geological materials

Both laboratory experiments and theoretical models have shown seismic attenuation to be sensitive to various petrophysical properties of rock and pore fluids, making attenuation a potentially useful time-lapse attribute to measure and interpret through rock physics models. Attenuation is rarely used as an interpretation tool due to the scarcity of reliable measurements and the wide choice of proposed mechanisms potentially responsible for the energy loss. Current methodologies for estimating attenuation from seismic data (spectral ratio, centroid frequencies and instantaneous frequencies) have been applied to the specific geometries associated with prestack surface seismic data and vertical seismic profiles, and adapted where possible for direct comparison of waveforms from two vintages of data. Phenomena that inhibit reliable estimation of Q can be repeatable between vintages of data allowing the true change in attenuation to be found, even when apparent attenuation is several times larger then intrinsic attenuation. Time-lapse attenuation measurements are sensitive to other waveform changes. Two examples of this are; errors due to changes in dispersion being introduced through frequency dependent travel-times (up to 40% error on 1/Q) and frequency dependent reflectivity, the latter being particularly sensitive at low frequencies and where polarity of reflections change with angle; band-limited random noise and multiple energy both decrease the estimate of attenuation (by 39% for a signal to multiple ratio of 2:1) but leave the change in attenuation detectable, with band-limited random noise also increasing the uncertainty. Attenuation measurements are then made on two field data-sets; a time lapse VSP from a pilot CO2 sequestration experiment where 1600 tonnes of CO2 was injected into a thin (10m) aquifer; and a prestack surface seismic data-set from a mature hydrocarbon reservoir undergoing enhanced oil recovery through alternating water and gas injection. For the VSP, attenuation changes are more detectable when directly comparing waveforms between vintages: however, the realistic synthetic (with a Q change of 100 to 20) indicated that the true magnitude of attenuation is unlikely to be recovered for such a small injection interval. In the real data, changes in attenuation (�(Q−1)=0.024) and velocity (5% decrease) are qualitatively interpreted by use of a patchy saturation model as an increase in CO2 saturation of between 10-30%. Four methodologies are used to calculate attenuation from the surface seismic data-set and show coherent anomalies indicating the robustness of the measurement. The region adjacent to a water and gas injector shows an increase in Q−1 of � 0.02. Changes in velocity (±5%) and amplitude (up to 150%) are also measured from the data and qualitatively agree with the attenuation measurements. A new quasi-linear inversion scheme is introduced to take these time-lapse attributes and solve for pressure and saturation changes using a patchy saturation model, giving 1-10% increases in gas saturation and up to 5MPa pressure changes around three water and gas injectors.