Store Carbon Underground: CCS from Site to Certainty
Storage is the petroleum system run in reverse: instead of finding what a trap kept, you must prove what it will keep. Screen the container, put the plume in motion, watch it move for decades, and turn monitoring into the certainty a regulator will sign.
You can screen a storage site with trap, seal, and pressure arguments, say which trapping mechanism is holding the CO2 at every timescale, prove a monitor survey will see the plume before the money is spent, process and read the 4D difference without fooling yourself, and turn caprock heterogeneity into the containment probability a permit actually asks for, with Sleipner as your reference case throughout.
The container
A storage site is a prospect evaluated backwards: the geometry that held buoyant fluid for a million years is the opening argument that it can hold CO2 for ten thousand. Container first, everything else after.
Injection is a pressure transaction. Gradients, contacts, and compartment boundaries set how much the site can take and where the displaced brine must go; measure them before the first tonne goes down.
Injection and the plume
A CO2 plume is a mobility contrast on the move: buoyant, thin, and eager to run ahead under the seal. Relative permeability and mobility ratio decide its shape long before any survey sees it.
Structural, residual, solubility, mineral: each mechanism takes over on a slower timescale than the last, and every containment claim you will ever file is a statement about which one is holding right now.
Injection raises pressure, and pressure moves stress: reactivated faults and a fractured caprock are how storage projects die. Coupled geomechanics is the difference between a capacity estimate and a safe injection rate.
Two limits cap the injection pressure: the caprock fractures at one number and the critically stressed fault slips at a far smaller one, and the safe ceiling is whichever comes first. With a well-oriented fault in the model, the pressure budget collapses from ten megapascals to half of one.
McGarr bounds the largest earthquake an injected volume can drive, and the bound climbs two thirds of a magnitude per decade of volume; a storage project injecting for decades must carry that arithmetic into its permit.
Watch it move
Supercritical CO2 into brine sand drops impedance over twenty percent, loud at any threshold, but the curve saturates after the first third: mapping the footprint is easy, quantifying the saturation inside it is the fizz problem in reverse.
Before a monitor survey is shot, fit-for-purpose modeling must answer whether the plume is even visible: Gassmann for the signal, repeatability noise for the floor. Detectability is a feasibility number, not a hope.
Two surveys shot years apart only subtract cleanly if processing treats them as one experiment; NRMS is the honesty metric that says whether the residual is noise or signal. For storage the difference is the deliverable.
Surface seismic sees the plume; boreholes hear the pressure. 3D VSP sharpens the image where the well is, and microseismic listens for the caprock starting to complain, the earliest warning storage has.
A storage permit outlives every acquisition crew that serves it. Sensors cemented in place make repeatability a property of the installation instead of a hope, which is what a decades-long obligation requires.
A difference volume is a map of what the site did since baseline, if you read it with the rock physics attached. For storage the reading becomes a containment statement someone must sign.
A 3 to 7 percent amplitude change must survive three decades of repeat surveys; joint processing with shared parameters is the only way it does. Walk the pipeline that turns eleven vintages into one defensible plume map.
Certainty
Containment fails at the thinnest cell, not the average one; the question is tail-driven, and realisations are how you count the tail. This is where site characterization becomes a number a permit can cite.
Only the plume changed between surveys: hold the baseline fixed and the inversion shrinks from thousands of unknowns to four. That prior is the production Sleipner recipe, and it is worth seeing it work.
The first industrial CO2 store and the longest-running public 4D record; every storage argument you make will be measured against it. Read the retrospective as the standard of evidence, not as history.
One project, every lesson: pick a rate the capacity and the seal can hold, inject, stop, and prove the secure fraction climbs. The deliverable is the defended number, not the pretty plot.