Induced seismicity has associated hazards and risks that can, in concept, be quantified. Understanding what is meant by “hazard” and “risk” related to induced seismicity is critical to any discussion of the options that can be employed to mitigate the possibility of felt induced seismicity and potential impacts from development of energy technologies. To promote a better understanding of hazards and risks, we first define these terms precisely and identify the factors that influence them. The remainder of the chapter discusses hazards and risks associated with induced seismicity and steps that can be taken to quantify hazard and risk associated with induced seismicity. The committee envisions future approaches toward mitigation of any hazards associated with induced seismicity involving “best practices” protocols as a cooperative endeavor between industry, government, and the public (Chapter 6).
The hazard of induced seismicity is the description and possible quantification of what physical effects could be generated by human activities associated with subsurface energy production or carbon capture and storage (CCS). For this discussion, physical effects include microseisms, earthquakes, and the associated ground shaking, both underground and at the Earth’s surface. In concept it is possible to calculate probabilities of the occurrence of microseisms and earthquakes and, given one of these events, to predict the possible ground motions. However, making such calculations requires assembling statistical data that are not readily available, such as the total number of wells of different depths, the geologic environments (including faults and plate motions), production characteristics from the well(s), and the subsets of those wells that generate microseisms and earthquakes of various magnitudes.
The risk of induced seismicity is the description and possible quantification of how induced earthquakes might cause losses (damage to structures, and effects on humans including injuries and deaths). The losses generally occur on the Earth’s surface, although
underground losses, for example damage to nearby petroleum wells, could also be analyzed. The concept of risk involves predicting the effect of induced ground motions, and perhaps fault slip, on structures and humans. If structures can incur moderate or heavy damage, risk involves predicting the effect of that damage (e.g., structural collapse) on humans in the vicinity.
Note that risk involves loss caused by structural damage, including effects on humans. If no structures or other constructed facilities are present, for example because the causative earthquakes occur in an uninhabited area, there is no risk. (Exceptions always exist to these general statements. One case would be an earthquake causing a rock slide that injures hikers in a national park, with no structure involved, but such cases would be rare exceptions.) The concept of risk could also be extended to include frequently occurring ground shaking that is a nuisance to humans (in the general, rather than legal, sense).
Factors Affecting Hazard
A set of questions can be addressed to understand and possibly quantify the hazard and risk associated with induced seismicity associated with energy technologies (Figure 5.1). Descriptions of each question are as follows:
1. Does an energy technology at a particular location generate apparent seismic events (meaning those that are felt at the surface)? The large majority of activities associated with hydrocarbon production do not cause any apparent seismic events. If no seismic events are recorded, this may be because the seismic events are too small (e.g., M < 0.0) to be recorded by regional seismic instruments, but the effect is the same: there is no apparent seismic activity.
2. Does an energy technology at a particular location generate just microseisms, or microseisms and earthquakes? This question involves the size of seismic events that are associated with the energy technology. Microseisms (by definition, seismic events with M < 2.0) generally do not produce ground motions strong enough to have an effect on structures, but they can in cases of close proximity be felt by humans at the surface. For example, two shallow (~2 km deep [1.2 mile deep]) seismic events of M 1.5 and M 2.3 in Blackpool, England, were reported by a number of people to have been felt in April and May 2011 (BGS, 2011).
3. Can earthquake shaking be felt at the surface? Not all earthquakes are felt at the surface. Earthquake ground motions at the surface depend on the size (magnitude) of the event and its depth, among other factors. The deeper the earthquake, the larger it must be to cause ground motions at the surface that can be felt by humans. Very shallow seismic activity (e.g., 2 km) has a higher hazard of causing felt ground motions than deeper activity.
4. What are the shaking effects? If ground motions are strong enough to be felt, they can be represented by three categories, depending on the maximum strength of shaking. Ground motions fall into the following categories:
a. Very slight shaking. These are felt ground motions, typically with peak accelerations less than 4%g, and do not cause damage to structures. Isolated cracks in plaster walls may be observed or items in houses may be knocked over, but these motions cause no damage of consequence. Frequent occurrence of these motions may be a nuisance to people.
b. Minor shaking. These are ground motions that frighten people and/or wake them from sleep, typically with peak accelerations between about 4%g and 18%g. If structures are present, these ground motions may cause light property damage (cracks in concrete, broken windows, or cosmetic damage) but do not cause buildings to collapse.
c. Moderate-strong shaking. These are moderate, strong, or severe ground motions with the potential of causing moderate or heavy damage, typically with peak accelerations greater than about 18%g. If structures are present, moderate to heavy damage may occur, including partial or complete collapse of structures or structural elements (foundations, walls, roofs). These effects on structures may cause human casualties (injuries and deaths, in severe cases).
Ground motions from induced seismicity generated at shallow depths can be more troublesome compared to the ground motions from deeper events (Figures 5.2 and 5.3). In a cross section of the Earth where a deep tectonic (natural) earthquake occurs at a depth of 10 km (Figure 5.2), the semicircles illustrate the distance within which minor shaking (or greater) occurs if the earthquake M is 3, 4, or 5. Because of the depth of the earthquake, minor (or greater) shaking usually does not reach the Earth’s surface for M 3 or 4. For M 5, minor (or greater) shaking may occur at the Earth’s surface within about 15 km (9 miles) of the epicenter (Figure 5.2).
Figure 5.3 shows a similar cross section of the Earth where a shallow earthquake occurs at the bottom of a 2-km-deep (1.2-mile-deep) well. Because of this shallow depth, a M 4 earthquake can cause minor (or greater) shaking within about 8 km of the well, and a M 3 earthquake may cause minor (or greater) shaking very close to the well.
Factors Affecting Risk
Risk from induced seismicity only occurs if structures are present that may be damaged. Risk exists to those structures only if the shaking is minor, moderate, or larger. Factors that should be considered for risk include location of faults, location of infrastructure that can be
FIGURE 5.2 Cross section of the Earth illustrating the maximum distance that minor (or greater) shaking will occur, for tectonic earthquakes originating at 10 km (6 miles) depth, with M 3 (green line), 4 (yellow line), and 5 (red line). In this example, only M 5 earthquakes will generate shaking that is felt at the surface.
FIGURE 5.3 Cross section of the Earth illustrating maximum distance that minor (or greater) shaking will occur, for both natural and induced earthquakes originating at 2 km (1.2 miles) depth, with M 3 (green line), 4 (yellow line), and 5 (red line). The diagram depicts an induced earthquake at the bottom of a well. Because of the shallow depth, each of these earthquake magnitudes would generate shaking at the surface that could be felt. Because of the larger energy released, a M 5 earthquake would be felt over a much greater area of the surface (up to 20 km [12 miles]) from the well, whereas a M 3 earthquake would only be felt about 1 km (0.6 miles) from the well.
damaged, and net changes to subsurface pore pressure caused by the energy project. These net changes involve the volume and pressure of fluids injected or extracted, the duration of injection and extraction, and the number of wells involved in the project. Note that these variables may be related; that is, the total fluid volume depends on the duration of injection or extraction and the number of wells involved.
Two spatial aspects of risk analysis are important to consider in the context of induced seismicity:
1. Multiple structures that can be damaged. A single well that induces earthquakes large enough to cause damage at the surface may damage multiple structures at the surface. If seismicity migrates during well operations (which is common for disposal wells), earthquakes have multiple opportunities to impact many structures. Even a small community located near a single well will have multiple structures with a range of vulnerabilities to ground shaking. Multiple structures give an increased chance of having one or a few structures with very weak resistance to ground shaking. Operations located in areas with many structures, such as the Basel, Switzerland, geothermal project, clearly have higher risk than a similar project in an unpopulated area. Likewise, CCS operations that are located at power plants in or near urban areas and which have the potential through injections of large amounts of CO2 over long time periods to increase reservoir pressures over large areas that may have surface developments may have increased risk.
2. Multiple well locations. The risk associated with induced seismicity has to be evaluated in terms of the sources of human activities. A geothermal operation, for example, may have multiple injection wells, each of which may generate seismic events that can affect different communities. For a large petroleum field, multiple wells may be used to inject fluid for secondary recovery, and each well may generate earthquakes that can affect separate communities. The spatial distribution for an entire industry project (e.g., underground injection of CO2) may be very large, and a risk analysis of the entire project would necessarily include that large spatial distribution and the multiple structures in that spatial area which induced seismic events might affect.
If a small number of wells (e.g., 10) are put in operation, the maximum shaking associated with earthquakes induced by those 10 wells can be described (Figure 5.4). In this example, a majority of wells (9 out of 10) will produce only felt motion, and only 1 out of 10 will produce ground motion with the potential for minor damage. No observations of moderate or greater (abbreviated hereafter as “moderate+”) damage occur in this example.
If many wells (e.g., 1,000) are put into operation, a histogram of the maximum shaking induced by those 1,000 wells would show that 250 wells are expected to produce ground motions capable of minor damage to structures. Ten wells are expected to produce ground motions capable of moderate+ damage to structures.
FIGURE 5.4 Example of relative probability distribution of maximum shaking at the ground surface from induced seismicity caused by one well. The relative probability increases upward on the vertical axis. The horizontal axis shows several kinds of measurements or effects of ground shaking: the upper scale indicates the amount of shaking (slight through moderate+); the second scale indicates ground acceleration, which increases from left to right; the next scale indicates MMI or the Modified Mercalli Intensity scale, which indicates the level of ground shaking at a particular location and has units designated by Roman numerals, also increasing from left to right in the level of ground shaking (see also Chapter 1); and the lower scale is the “felt” effect, ranging from “felt only” on the left through minor to moderate or greater (“moderate+”) damage. The probability of very slight shaking is much higher than for moderate+ shaking (or damage) for one well that causes an induced seismic event of any magnitude.
A more general distribution of ground motion from a range of earthquakes with ground motions quantified by the largest horizontal acceleration1 that occurs shows that the majority of shaking will be in the category of “felt only” (Figure 5.4). A small percentage (~25 percent) may have the potential to cause minor damage, and a very small percentage (~1 percent) may have the potential to cause moderate+ damage (Figure 5.4).
The important conclusion is that, while the risk of minor, moderate, or heavy damage from induced earthquake shaking may be small for each individual well, a large, spatially distributed operation leads to a higher probability of such damage. If we define PM as the probability of moderate+ damage given surface ground motion from one well, then the prob-
1 The peak horizontal acceleration of the ground is a common measure of ground shaking because the maximum force on objects sitting on the ground is proportional to the peak horizontal acceleration through Newton’s second law. Acceleration is measured in units of gravity, “g,” which is the acceleration of a falling object. For comparative purposes, a modern, high-powered sports car can accelerate at about 50%g.
TABLE 5.1 Probability of Damage Increases with Number of Wells
|Total Number of Wells (N)||[PM]N wells||Expected Number of Wells Causing Moderate+ Damage|
ability of at least one observation of moderate+ damage given that N wells are in operation can be calculated2 as
This probability increases with the number of wells N (for PM = 1%), as shown in Table 5.1.
This example illustrates that, as an industry begins operation with a few wells, there might be no apparent problem with induced seismicity. As the industry expands to 100, 1,000, or more wells, there can be a significant likelihood that induced seismicity will cause damage to structures somewhere, as a result of the large number of earthquakes and ground motions that are induced, even though the probability of any one well producing such ground motions is small.
Tectonic earthquakes cause some level of earthquake risk for buildings, primarily in areas like California with relatively frequent events. Seismic building codes provide some level of protection but are not a guarantee against earthquake damage. In other regions, building codes provide lower levels of seismic protection, and earthquakes (whether tectonic or induced) may cause damage, depending on the level of ground motion associated with them.
Several steps can be taken to quantify hazard and risk. As described in the previous section, the quantification of hazard and risk requires probability assessments, which may be
2 This is a special case of the Bernoulli distribution with N independent trials and probability PM of occurrence of the phenomenon of interest (moderate+ damage). The probability of at least one observation of this damage is 1 minus the probability of no observations of this damage, given N independent trials. Any dependence among ground motions for a given technology can be examined as part of the hazard assessment step identified in the section Quantifying Hazard and Risk (this chapter), in particular step 3 in Table 5.2.
either statistical (based on data) or analytical (based on scientific and engineering models). Thus, for implementation, some of the steps will require the collection of statistical data. Other steps can modify and use analytical models that have been developed for hazard and risk analysis of tectonic earthquakes. Table 5.2 summarizes the steps that can be taken to quantify the hazard and risk of induced seismicity for a single project (a single wastewater disposal well, oil or gas extraction well, etc.).
Step 1 in Table 5.2 involves estimating the probability of generating earthquakes with M ≥ 2.0. This is a statistical problem that can be addressed only by collecting statistical data on the number of wells drilled for each technology, their characteristics (depth, volumes of fluids, pressures, rates of injection or extraction), and observations on whether they generate earthquakes. Simulation models that predict fluid flow in the Earth’s crust given characteristics such as permeability, pumping rate and volume versus time, geologic units (including ages of the rocks), and other factors can be the basis for predictive models, and these models can be refined on a probabilistic basis as more data and observations are gathered and analyzed. The cell labeled “1C” in Table 5.2 indicates that these statistics will be technology dependent, because the typical volume of fluid, pressure at which it is injected, and other factors in a given project depend on the energy technology. Cell 1D indicates that energy projects in tectonically active regions can be expected to have a higher probability of generating M ≥ 2.0 earthquakes than do energy projects in tectonically stable regions. Finally, cell 1E indicates that the probability assessment from statistics will have a depth dependence: large earthquakes are less likely to be induced by shallow wells.
Step 2 involves estimating the probability of felt shaking at the surface (see cell 2A, Table 5.2). This is an analytical problem with some statistical inputs (cell 2B). Specifically, data are needed on the frequencies of occurrence of different earthquake magnitudes. As cell 2C indicates, these frequencies are expected to be technology dependent. The reason is that, among energy technologies, earthquake-generation mechanisms vary (Chapter 2), and the net injected or extracted fluid volume varies (Chapter 3). Once these data on magnitude distributions are obtained, analytical methods are available to estimate shaking (see, for example, Boore, 2003). This probability may be region dependent (cell 2D) because earthquakes in stable crustal regions may release higher levels of crustal stress than similar-magnitude events in active crustal regions. Finally, the probability of felt shaking will depend on the depth of the induced earthquakes (cell 2E) (see Figures 5.2 and 5.3).
Step 3 involves estimating the probability of different strengths of earthquake shaking (cell 3A). This is a well-studied problem in seismic hazard analysis for tectonic earthquakes, for which analytical techniques are available (cell 3B). The result will depend on energy technology (cell 3C) because observations of earthquake magnitude distributions, particularly the maximum magnitude, have some dependence on energy technology (see Figure 3.15). Also, the result will depend on region (cell 3D) and depth (cell 3E), because earthquake magnitude distributions depend on these factors.
TABLE 5.2 Steps for Hazard and Risk Assessment for a Single Project
|Step (see corresponding box in Figure 5.1)||A. Probability needed||B. Method||C. Technology Dependent?||D. Region Dependent?||E. Depth Dependent?|
|1||1A. P[generate M ≥ 2 earthquakes]||1B. Statistical||1C. Yes, depends on factors such as volume, pressure, rate, and depth||1D. Yes, tectonically active versus stable region||1E. Yes, large earthquakes usually not induced near surface|
|2||2A. P[shaking felt at surface]||2B. Analytical/ Statistical||2C. Yes, depends on magnitude distribution and maximum magnitude||2D. Yes, depends on earthquake properties||2E. Yes, deeper induced earthquakes may not be felt|
|3||3A. P[strength of shaking]||3B. Analytical||3C. Yes, depends on maximum magnitude||3D. Yes, depends on earthquake properties||3E. Yes, shallow earthquakes will generate stronger shaking|
|4||4A. P[structures and people affected]||4B. Analytical||4C. No||4D. Yes, depends on structural strength and tolerance for shaking||4E. Yes, deeper earthquakes, if felt at the surface, may affect a larger area|
NOTE: Gray shaded cells indicate methods that have to be developed to estimate probabilities (“P”) for various aspects of an induced seismic event shown in the green-shaded cells. These four aspects include the probability of generating an earthquake of M > 2.0, the probability of shaking being felt at the surface, the probability of different strengths of shaking from an earthquake, and the probability that the earthquake shaking will affect structures and people.
Finally, step 4 involves estimating the probability that structures and people are affected (cell 4A). Analytical methods for seismic risk analysis (cell 4B) are well established for tectonic earthquakes, and these should be applicable to induced earthquakes. The methods will not depend on technology (cell 4C), because a structure’s response does not depend on how the shaking was generated. However, the methods do depend on region (cell 4D); structures outside of California and Alaska are generally not designed to withstand high levels of ground shaking, and people in aseismic regions may be less tolerant of low-level shaking than those who have previously felt natural earthquakes. Deeper earthquakes will have an influence on the numbers of structures and people affected (cell 4E) if the associated earthquake shaking covers a wide region and affects more structures and people.
Table 5.2 summarizes steps that can be taken to estimate hazard and risk for individual energy projects. The specific statistical data that need to be collected, and analytical methods that need to be modified from other fields, are summarized in column B. Each of the statistical or analytical methods in column B will calculate the probability indicated in the corresponding cell in column A, and these calculations will depend on the corresponding cells in columns C, D, and E. For instance, statistical data on M ≥ 2 earthquake generation (cell 1B) need to be collected and analyzed by energy technology, volume of fluid, injection pressure, rate of injection, etc. An unstated assumption in Table 5.2 is that data are to be collected for new energy projects in areas that are known to have a history of induced seismicity, as well as existing projects. The reason is that, going forward, we presumably are interested in estimating hazard and risk from induced seismicity caused by further expansion of energy production, not by existing energy production. However, data from existing projects will allow forecasts of induced seismicity for industries as a whole. The distinction is important: seismicity induced by a new injection or disposal well will differ from seismicity induced by a well that has been in production for years, where crustal stresses may have equilibrated.
Note that steps 1 through 3 apply regardless of whether the potential induced seismicity will occur in areas of high population or sparse population. Step 4 determines the effect on structures and people, and this effect of course depends on the location with respect to structures at risk and people. Induced seismicity could be caused in a region of sparse population, affecting few people, but could affect dams, bridges, or power plants, with large concurrent costs.
These steps, if developed, can be used in three important ways:
First, by compiling statistics on earthquake generation by technology and characteristics (cell 1C), insight can be gained on what combinations of volumes, pressures, rates of injection/extraction, and so on lead to higher probabilities of induced seismicity. This insight can be used to create well-documented, data-based input to best practices protocols (see also Chapter 6).
Second, energy technology development, whether through public or private efforts, will have data with which to make decisions to minimize induced seismicity effects on
people and structures. For example, if a particular project is observed to generate M ≥ 2 earthquakes (i.e., the probability in cell 1A becomes 1 for that project), decisions can be made on pumping characteristics to minimize the probabilities of shaking felt at the surface (cell 2A) and of strong shaking (cell 3A).
Third, the calculated probabilities of shaking felt at the surface (cell 2A), of strong shaking (cell 3A), and of structures and people being affected (cell 4A) can be generalized from those for one project (as depicted in Table 5.2) to forecast the total number of induced seismicity cases that will occur and the number of structures and people affected. If detailed statistical data can be obtained for cells 1B and 2B, this generalization can account for details on forecast locations of projects, volumes and other characteristics of pumping, and proximity to inhabited areas. The estimated numbers of people and structures affected can then become the basis for decisions on whether and how to minimize the impacts of induced seismicity.
Directed research could support development of these steps for the quantification of hazard and risk, with the overall goal of integrating these steps to improve our capability to predict induced events and their consequences. Chapter 6 develops these ideas further by discussing best practices and protocols to avoid or mitigate the impacts of induced seismicity during energy development projects.
BGS (British Geological Survey). 2011. Blackpool earthquake, Magnitude 1.5, 27 May 2011. Available at www.bgs.ac.uk/research/earthquakes/blackpoolMay2011.html (accessed November 2011).
Boore, D.M. 2003. Simulation of ground motion using the stochastic method. Pure and Applied Geophysics 160:635-676.