Learnings From the SEG/SPE/ARMA Injection-Induced Seismicity Workshop
Shawn Maxwell, Workshop Chair
In September, a technical workshop was held on the topic of injection-induced seismicity in Banff, Canada. It brought together industry and technical experts to discuss the increasingly important topic of induced seismicity associated with various injections during oil and gas activities.
The event was cosponsored by the Society of Exploration Geophysicists (SEG), the Society of Petroleum Engineers (SPE), and the American Rock Mechanics Association (ARMA), serving as a follow-up to a previous meeting on the topic held in Broomfield, Colorado, in 2012.
More than 120 professionals participated, with the majority traveling from the United States (60%) and Canada (30%), and some from Europe (10%) and one each from Japan and Colombia. The attendees represented oil and gas companies (34%), academia (25%), service companies (25%), and government organizations including national laboratories, geological surveys, and regulatory bodies. Similar to the workshop in Broomfield, the majority of the participants were geophysicists (both industry geophysicists and earthquake seismologists) with engineers and geomechanicists making up the rest of the group.
Nearly 50 technical presentations made up the program that covered case studies, various techniques to model the phenomena, seismic hazard and risk, and industry protocols to mitigate the problem. The workshop was organized through a committee consisting of Dave Eaton, University of Calgary; Hal Maccartney, Pioneer Resources; Julie Shemeta, MEQ Geo; Mike Fehler, Massachusetts Institute of Technology; Norm Warpinski, Pinnacle/Halliburton; Pat McLellan, Talisman Energy; Rod Gertson, Devon Energy; and was chaired by Shawn Maxwell, IMaGE.
It was interesting to compare and contrast the two workshops to see how experience, understanding, and attitudes have changed during the period. The Broomfield meeting was held at a time when concerns about injection-induced seismicity were just beginning and the meeting was primarily intended to educate the industry about the issue. As described in a Broomfield workshop report published in SEG’s The Leading Edge and SPE’s Journal of Petroleum Technology, key facts were established.
When Does Seismicity Occur?
For example, only a small proportion of injection wells have induced seismicity, including geothermal, waste injection, and hydraulic fracturing operations. The seismicity occurs when preferentially oriented faults close to the point of slipping in the geologic stress field have stress and/or pressure changes that cause the fault to slip. Seismicity also can occur after the injection has ceased and may occur at some distance from the actual injection point, which can lead to uncertainty identifying injection as a causal factor.
Traffic light systems were described as a basis for operational protocols, where operational changes are made in reaction to a seismic response causing a “yellow” warning light, or ultimately stopped if conditions lead to a “red” light. Experience gained since 2012 was evident in the presentations and technical discussion, with improved knowledge in certain areas while substantial technical uncertainty remained in others.
Since the Broomfield workshop, additional occurrences of induced seismicity have occurred, notably associated with waste injection (particularly in Oklahoma), hydraulic fracturing (in western Canada and Ohio), and geothermal and gas production and storage in Europe. Seismic activity in Oklahoma dramatically increased in the first part of 2014. Rising concerns have kept the topic of induced seismicity on the radar screen of industry practitioners, scientists, and regulators. Technical discussions in Banff carried both a tone of timeliness and a sense of urgency about the issue.
Although more examples have come to light since the Broomfield meeting, it was also apparent that more sophisticated hydraulic, geomechanical, and seismic modeling tools are being accessed. Similarly, industry mitigation protocols are being implemented, largely involving the so-called traffic light system based on observed magnitudes determined from dedicated monitoring arrays.
One issue raised at the Broomfield workshop resurfaced at Banff regarding definitions of common terminology. Most experts seem to agree that the primary mechanism of felt earthquakes is associated with injection-elevating pressure on tectonically stressed faults already close to the point of slipping. The implication is that the injection triggers the tectonic stress release of the fault. In contrast, an induced slip scenario would result from increases that are significant enough to elevate conditions all the way to failure.
Although the mechanism and related tectonic settings are very different between triggered and induced seismicity, the two terms are generally used interchangeably. Triggered seismicity also has other meanings in seismology, for example, aftershocks are mechanistically described as being triggered by the main shock or remote triggering of local seismicity after large earthquakes. Perhaps “injection-pressure, triggered seismicity” would be more descriptive terminology. Nevertheless, “induced seismicity” has become the colloquial term (as used in the title of the workshop and throughout this report) regardless of the mechanism.
The other potential point of confusion is the use of the term “microseismicity” to describe background seismicity associated with felt seismicity, in many cases with magnitudes much greater than zero and hence much larger than typically detected during the common practice of microseismic imaging of hydraulic fracturing. Perhaps the term “microearthquakes” would be a better industry term describing the background seismicity and to avoid confusion with the lower-level activity typically used to image hydraulic fracturing and other reservoir processes.
Another issue common to the Broomfield discussions is the continued difficulty in establishing cause and effect between a felt earthquake and local injection operations. A classic paper by Davis and Frohlich (1993) is often cited as providing accepted criteria to distinguish induced from natural seismicity. However, the described criteria focus on seismicity correlating spatially and temporally with a single injection, whereas there is an emerging notion that there may also be a more regional consequence resulting from multiple injections within an area.
A number of cases of suspected induced seismicity also occur after the end of injection and at distances and depths away from the injection point. The inference is that a pore pressure plume continues to diffuse through permeable rock with time after the injection pressure has stopped and can migrate outward and downward from the original injection point. In fact, many case studies of suspected induced seismicity presented in Banff included a relatively simplistic (homogeneous properties) pore-pressure model.
Although there are some efforts with more sophisticated flow models, few compare with the level of sophistication of petroleum reservoir flow simulations common in reservoir management. In contrast, it is also difficult to discount a causal relationship with injection so that natural seismicity occurring coincidentally near an injection well is often assumed to be induced.
An important discussion topic that emerged in Banff was the lack of a validated model of ground motion for induced seismicity. Such ground motion models are a fundamental component of probabilistic seismic hazard assessment, which is an analysis tool used, for example, to develop building codes in tectonic earthquake regions. Induced seismicity tends to be shallower and possibly has a different frequency content and perhaps source mechanism, such that the characteristics of the ground motion close to the earthquake may be distinct from common models used for natural seismic hazard. Indeed, the limited observations of ground motion at close distances to induced seismicity appear to vary from the typical models used for natural seismic hazards.
However, a significant statistical population of observations in a widely varying range of geological environments does not exist. Seismic hazard and the associated risk potential are the key concern with induced seismicity, and the lack of ground motion models is a fundamental gap in knowledge, thus leading researchers to plead for released data particularly at close distances to assist in ground motion calibration.
A related topic is the impact of induced seismicity on national seismic hazard maps, which are directly tied to regional building codes. These seismic hazard maps do not include any component of hazard associated with induced seismicity, and government bodies responsible for the assessment, such as the US Geological Survey, are currently considering how best to make appropriate updates.
Another important discussion topic was the need for the creation of industrial best practices: how best to access site conditions, engineer the injection, and react if induced seismicity becomes a problem. Workshop participants concluded that a “one-size-fits-all” practice would not suffice and that the solution needs to be tailored to local geomechanical and tectonic conditions, injection and operational characteristics, and hazard and risk implications in the region. Various existing industry protocols with proactive measures to establish a way for the industry to move forward and mitigate the seismic risk were also presented.
Nevertheless, the industry continues to be taken by surprise as suspected incidents of induced seismicity first come to light in a region, thus suggesting a need for better site screening. Once the problem surfaces, monitoring efforts tend to increase and become reliant on local seismic monitoring. The resulting operational procedures tend to gravitate to a traffic light monitoring system using magnitude levels to define the mitigation thresholds.
However, the workshop discussions highlighted the variability in setting these levels, depending on a variety of local factors. The issue of which parameters are best used to trigger traffic light changes was raised, for example, estimated source magnitude versus perhaps the resulting ground motion levels. The effectiveness of traffic light systems was also questioned, particularly around the success of the system in dealing with induced seismicity in European geothermal projects.
There are few examples where such systems have been successfully used to modify an operation to mitigate seismicity. There is also uncertainty about how best to change the operation to react to changing warnings or even a clear definition of success in mitigating seismic risk. Regardless, it is clear that magnitude-based traffic light monitoring systems have become the primary tool that the industry is using to move forward.
Many researchers also repeatedly raised the issue of access to quality data, a clear necessity to advance the science. Larger events detected on national arrays are available through open databases, enabling researchers to develop new processing techniques such as signal matching to revisit data sets to find weaker, undetected background seismicity. In many cases, companies are installing their own proprietary local arrays, which provide unique and much more sensitive recording to obtain a more complete seismic data set. Even in the case of open data sets, there is a clear need for complimentary reservoir engineering data including reliable records of injection volumes, rates, and pressures that are often lacking. Long-term data related to stress and strains before, during, and after injection are also a missing pieces of the puzzle.
Finally, one critically important topic was raised on how to engage and educate the public about the issue, although in reality it extends past the scope of a technology-based industry workshop. While participants agreed on the importance of a transparent and open approach with the public and the media, including explaining both what we know and do not know about the issue, it is unclear whose shoulders this falls on.
The success of the Banff workshop is a direct result of the willingness of presenters to share their technical expertise, the attendees who shared a vast range of experiences, and of course, the hard work of the organizing committee. I am profoundly thankful to all who assembled in Banff and for their respective contributions, combining to make Banff a memorable and landmark meeting in the advancement of this critical topic. I keenly look forward to the next meeting in this series to see the progression over the next 2 years.
Davis, S.D. and Frohlich, C. 1993. Did (Or Will) Fluid Injection Cause Earthquakes? Seismological Research Letters 64 (3-4):207-224.