CO2 Plume Time-Lapse
Learning objectives
- Understand 4D as a baseline-minus-monitor difference
- See the plume appear only after the static geology cancels
- Weigh the 4D signal against repeatability noise
- Treat modelling as a detectability feasibility test
The Brief
CO2 is being injected into a saline aquifer for permanent storage, and regulators demand proof it stays contained. A baseline survey was shot before injection; a monitor survey will be shot after. Find the plume, and, crucially, prove you can find it before the money is spent on the monitor.
The Build
Injecting CO2 lowers the reservoir impedance, replacing brine with a lighter, more compressible fluid, so the monitor section differs from the baseline only inside the plume. Subtract the two surveys and the static geology, identical in both, cancels exactly, leaving the 4D difference. In it the plume should stand out as a bright anomaly. But every survey carries repeatability noise, and that noise does not cancel. Whether you can claim a plume comes down to whether its signal rises above that residual noise.
The Debrief
Which engine? Convolution, plus Gassmann, judged against noise. The amplitude change is a zero-offset impedance effect that convolution renders perfectly well, so you do not need the wave equation. But you must model the fluid substitution to have any 4D signal to detect, and you must compare that signal to a realistic repeatability noise floor. Raise the saturation and the anomaly emerges; raise the noise and it drowns even though the plume is physically there.
That reframes what the model is for. The deliverable here is not a pretty image; it is a yes-or-no feasibility answer: given the expected saturation and the achievable acquisition repeatability, will the monitor survey actually see the plume? Fit-for-purpose modelling has become risk management. The final capstone reads a gas cloud whose slow velocity distorts time itself.