We present a new methodology to integrate 4D seismic data acquired continuously over three years to monitor gas injected in a land carbonate reservoir. A rock physics model is first calibrated from well log data, which is then used to transform reservoir simulation results to 4D synthetic seismic data. We show from this modeling that maximum 4D amplitudes of 6-10% NRMS are strongly impacted by gas thickness, while 4D time-shifts have more complex behavior and cannot be integrated in a simple way into a predictive model. Due to relatively high noise levels in comparison to the expected 4D signal, we show that quantitative models fail to be useful in this situation. Instead, we define a gas detectability threshold from a statistical model between gas thickness and 4D amplitudes, and calibrate a qualitative model of gas detectability depending on noise. This model is used to update in a Bayesian way prior detectability probability maps from the reservoir model, and finally, to update probabilistic gas volume maps from the reservoir model.
Presentation Date: Monday, October 12, 2020
Session Start Time: 1:50 PM
Presentation Time: 4:20 PM
Location: 362D
Presentation Type: Oral
