by Susan Hovorka
A lot of work has been done on designing monitoring programs for carbon capture and storage sites. All of the regulations, very properly, say that monitoring should be site-specific. But the details of how a regulator and an operator determine what is site-specific have not been fully explored. This creates uncertainty. What the site-specific phrase means to a regulator may not match up with what the site-specific phrase means to a site developer.
At the Gulf Coast Carbon Center, we have considered the ways that monitoring tools interact with sites. In this context, we have found it useful to think of these tools as traps for catching carbon dioxide leaks. Leakage from a well-characterized storage reservoir is not expected, however even from a site for which the characterization is excellent, some uncertainty remains. Stakeholders, such as regulators, capture industries, project financiers, or the public may find such uncertainty unacceptable. To borrow an analogy from a business whose entire goal is the elimination of the unacceptable: You can’t catch a mouse with a squirrel trap. You also won’t catch a mouse in a lake in the winter. You have to set the right kind of trap, in the right place, at the right time to determine if you do or do not have mice.
We explore how to set the right trap to catch leakage using four common tools as examples.
Time-lapse 3D seismic can be an ideal tool to detect leakage, because if CO2 migrates out of the reservoir and replaces brine, the resulting change in seismic velocity and amplitude can be identified by subtracting the survey parameters before leakage from the survey after leakage. The difference can be interpreted as a result of fluid substitution. However, at settings where the rock frame is stiff, the fluid substitution has a small impact on seismic properties. Because some other differences inevitably occur between the before and after surveys to cause non-repeatability and noise, a small signal can be obscured causing leakage detection to fail. It is therefore a best practice to 1) calculate the signal resulting from fluid substitution and 2) measure the noise before relying on time-lapse 3D for leakage detection. Such a calculation is normally done for designing a survey to detect changes in the reservoir where they are expected; it is even more important to use a robust design to show no change in area where no change is expected.
In many settings, groundwater geochemistry is an important parameter for reducing risk of groundwater contamination. For example, at a landfill or a chemical spill, an array of wells can be used to provide assurance that mitigation is effective and contaminants are not migrating into the regional groundwater. The same approach has been proposed to provide assurance the injected CO2 is retained and has not migrated to groundwater. However, monitoring groundwater for injected CO2 is different than monitoring for a known contaminant for several reasons. 1) Only in case of an unexpected failure would any migration occur. Groundwater monitoring is therefore mostly a matter of seeking to prove a negative. 2) CO2 is naturally present in groundwater. 3) CO2 is reactive with the rock-water system. Our work shows that rock properties are quite important in determining which geochemical properties are sensitive indicators of leakage. For example, in an aquifer that contains no carbonate, pH may be a reliable indicator of addition of CO2. In aquifers with some carbonate, bicarbonate is the indicator of the addition of CO2. Dissolved inorganic carbon (DIC) is the most consistent indicator of increased CO2.
One classic leakage indicator is the transport of hotter fluids from depth, which provides a thermal signal of leakage. The reliability of this tool is shown to depend on location of the potential leakage pathway as well as the leakage rate. A thermal signal is especially diagnostic in a CO2 storage setting. Before CO2 reaches the leakage pathway, brine leakage into an overlying unit produces a warming signal. As the leakage path begins to transport CO2, Joule-Thompson cooling as a result of expansion during pressure decrease causes a strong cooling trend. However, the mass of rock and fluid attenuates the thermal signal over short distances, so this method is not appropriate for reconnaissance over an area.
Above zone pressure monitoring is another powerful technique for documenting no leakage. However, to document no leakage, above zone monitoring must be designed with wells spacing adequate to be sensitive to leakage rate and leakage volume at the relevant hydrodynamic conditions.
Given these benefits and drawbacks of the tools available to catch feral carbon dioxide, how do we understand them in a site-specific context?
We expect our measurements to fall in a predicted range with an associated uncertainty. We make observations, and these observations also have associated uncertainty. The mismatch between measured and predicted values is commonly used to determine if an injection is behaving as expected. However, this approach can lead to inappropriate responses. A mismatch may make no material difference to the outcome of the projects. Further, the prediction and observation may not be designed to be sensitive to unacceptable outcomes. We suggest instead that the project define an acceptable range of outcomes. Any monitoring measurements or trends that fall inside the acceptable range show that the project is adequately conformant. In contrast, an observation outside the acceptable range, indicates a possible material impact and requires follow up.
A monitoring plan should begin by modeling site-specific and project-specific material impacts. This could include a trend toward unacceptable pressure increase or lateral migration rates that might allow the plume to become larger than acceptable to the project. Unexpected low pressure might indicate that fluids are migrating out of the intended zone. With a quantitative signal based on modeling the acceptable outcome, a monitoring plan can place the right tools at the right place to detect the unacceptable response. In many cases, a strong monitoring design is effective in reducing uncertainties to the point that monitoring can stop. In the mousetrap analogy, if we don’t catch anything, we can take down the trap. And we can stop worrying about the varmints.
We are grateful to the organizations the provided support for this research. Some elements of the project were funded by the Carbon Capture Project and some elements were funded by the US EPA.