Deep Well Injection Induced Seismicity

: Injection of fluid into subsurface geologic strata for geothermal energy, oil production, and waste disposal has been linked to induced seismic activity in the United States as well as in several other countries. According to the report of the National Research Council of United States of America thousands of induced earthquakes were reported at the numerous sites, where oil and gas recovery and waste disposal activities took place. Most of these induced earthquakes were small magnitude events (Moment Magnitude [M w ] < 4), although earthquakes of magnitude (M w ) 6.5 to 7 were also reported near the oil and gas production sites. This paper presents the results of a review of case histories on increased seismic events due to deep well injection (DWI) and oil extraction. Key factors that may lead or contribute to increased seismicity will also be discussed.

Increased seismic events caused by human activities have been reported and documented. Most of these seismic events were low magnitude events. The documented cases are typically related to energy-and oil-production activities and injection of waste for disposal, although other human activities, such as mining and reservoir filling, have also been shown to cause increased seismic events [1]. Table 1 summarizes the cases of induced seismicity reported for the various energy technologies, as of 2012 [2,3]. The table also lists the number of felt events and the maximum earthquake magnitudes recorded at these sites, including an earthquake of magnitude 6.5 observed in a hydrocarbon withdrawal project.
This paper focuses on case histories of increased or induced seismic events due to DWI of large-volume waste fluid. Cases involving induced earthquakes due to oil production were also reviewed and presented. Six of these case histories are discussed below.

Case 1: Paradox Valley Brine Deep Well Extraction and Injection, Colorado, USA
The facility includes, among others, 9 shallow extraction wells, a United States Environmental Protection Agency (EPA) Class V deep injection well (total depth of ±4.9 km below ground surface), and the Paradox Valley Seismic Network (PVSN). The seismic network consists of 20 stations of velocity sensors and strong-motion accelerometers installed around the injection well location. The shallow wells extract brine from the aquifer along the Dolores River in southwestern Colorado, and after some treatment, the brine is injected back with high pressure to depths between 4.3 and 4.8 km [4].
The primary target of injection is the highly fractured Leadville Limestone, where the steeply dipping Wray Mesa Fault system trends sub-parallel to the strike of the Paradox Valley. The injection well was sited to optimize fluid disposal along these fault fractures. From 1991 to 2003, more than 4 million cubic meters (about 1 billion gallons) of brine was injected into the rock strata, with an average injection rate of 855 to 1,290 liters per minute (l/min) or 0.325 to 0.49 million gallons per day (mgd) during the operational period. The area had experienced very low seismicity prior to the injection. After the injection activities began, more than 4,600 induced earthquakes were recorded [5]. These induced seismic events were observed in two distinct zones: a principal zone surrounding the injection well and a secondary zone centered at ±8 km northwest of the well location.
The injection and observations of induced seismic events at the PVU site can be divided into two main periods: injection tests and operational (continuous) injection. The following subsections summarize the well operation and induced seismicity observed during each of these periods [4].

Injection Testing Period (July 1991 to April 1995)
A total of seven well tests were performed during the permitting period to obtain the EPA Class V permit for brine disposal. The injection duration was varied, ranging from 12 days to as long as 8 months, with different injection rates and well pressure. A total of 666 induced seismic events were detected during this period, and they were observed to be correlated to injection rate and pressure.

Injection Operational Period (May 1996 to 2003)
During this operational period, the injection rate was slowly increased to a fixed rate of 1,290 l/min or 0.49 mgd, and the maximum pressure at the head of the well was capped at ±33 MPa. Induced seismic events were first detected after 111 days of continuous pumping, and more than 3,350 seismic events were recorded within ±10 km of the well location through the end of 2003. The majority of these events were  [2] and The National Academy of Sciences (2012) [3] small magnitude earthquakes (Mw ≤ 2.5); only 15 events were felt on the ground surface. The largest induced seismic events were two Mw 3.5 to 3. The operational (or continuous) period is further grouped into four phases, as summarized in Table 2 along with the injection parameters implemented during each of these phases. Figure 1 shows plots of injection rate and frequency of induced seismic events for the four phases of operation described in Table 2. The data indicate the following: • The induced seismic events were the highest in Phase I (through middle of 1999). • After the Mw 3+ earthquakes in 1999, PVU implemented a 20-day shutdown for every 6 months of continuous operation (Phase II). The purpose of these shutdowns was to allow the pressure at depths to diffuse, reducing the potential of inducing large magnitude earthquakes. As shown in Figure 1, these regular shutdowns significantly reduced the seismic activity (from as high as ±150 events per month to less than ±50 events per month). • Despite the lower seismic activity due to shutdowns, an Mw 4.3 event occurred on May 27, 2000. This event required the PVU to reduce the injection rate from 1,290 l/min to 855 l/min, a 33 percent reduction (Phase III). The reduced injection rate resulted in further reduction in induced seismic events (see Figure 1). • During Phase IV, the injectate composition was changed from 70 percent brine (PVB) plus 30 percent fresh water to 100 percent PVB, while the injection rate was kept at 855 l/min. This was done to increase the disposal rate of brine. As of the end of 2003, no noticeable increase (as compared to Phase III) in induced seismicity was observed. Figure 2 depicts the more recent data (through January 2011) collected by the Bureau of Reclamation [6,7]. These recent data confirm the observations and findings of previous studies and identify four key parameters for induced seismicity: injection volume, injection rate, downhole pressure, and percent of day injecting. Of these four parameters, downhole pressure exhibits the best correlation with the occurrences of near-well seismicity over time [6].

Case 2: Rocky Mountain Arsenal, Colorado, USA
The Rocky Mountain Arsenal site was used by the U.S. Army to manufacture weapons. In 1961, a well was drilled to a depth of ±3.7 km into the crystalline  [4] rocks for chemical fluid disposal. Small earthquakes were detected soon after the injection, with the majority of these events occurred within 5 miles of the injection point and aligned with the orientation of vertical fractures found in the rock. As reported by McClain (1970) [8], "in June of 1962, several earthquakes occurred which were large enough to be felt by residents and caused considerable concern. By November of 1965, over 700 shocks had been recorded and, although 75 of these had been felt, no damage was reported".

Case 3: Geysers Geothermal Steam Field, California, USA
The Geysers geothermal energy site is situated about 75 miles north of San Francisco in Northern California. It is one of the most productive geothermal fields in the world and has well-documented records of seismicity associated with geothermal energy development. The power plant was supplied with steam from a total of 420 production wells. After the steam was used to operate turbines for power generation, the waste was then injected back into the ground at similar depths using 20 injection wells. Because the volume of injected waste was less than the stream produced, water was later added to replace the mass loss.
The development area is in a relatively high seismic region near the San Andreas Fault system in Northern California, and no active faults are known to cross the site. Thousands of small earthquakes were detected soon after the steam production, and some of them caused damages to nearby buildings.    Evans (1966) [10], Healy et al. (1968) [11], McClain (1970) [8] and Hsieh and Bredehoeft (1981) [12]

Case 4: Montebello Oil Production, Southern California, USA
The 1987 Mw 6.0 Whittier Narrows earthquake occurred in the San Gabriel Valley of Southern California. This earthquake occurred on a blind thrust fault (i.e., the rupture did not extend to the ground surface) near the northern part of the Elsinore Fault Zone beneath an active oil production field at a depth of ±9.5 km below ground surface [16,1].
Although a causal relationship between the earthquake and oil-production activities can be considered weak due mainly to the ±8 km vertical separation between the earthquake hypocenter and oil-producing formation, a mechanical connection between the two has been postulated [16,1]. It is suggested that removal of oil and water from the upper crust may result in imbalanced forces in the deeper seismogenic layer that, in turn, induce earthquakes (see Figure 6 below).
This hypothesis is further supported by similar earthquake events in California (the 1983 Mw 6.5 Coalinga and 1985 Mw 6.1 Kettleman North Dome earthquakes) and in Gazli, Uzbekistan (the 1976 Mw 7.0 Gazli earthquake). These three earthquakes all occurred beneath oil-production fields.  Preiss et al. (1996) [13] and Beall et al. (1999) [14] Mcgarr, 1991[16] Although oil extraction is different than deep brine or wastewater injection, these case histories may contribute to the understanding of the potential for inducing moderate to large magnitude earthquakes by fluid injection in a high seismic tectonic plate boundary area, especially near an active fault system.

Case 5: Inglewood Oil Production , Southern California, USA
The failure of Baldwin Dam in 1963 has been in part blamed on the nearby Inglewood oil production activities. At the time of the dam failure, the Inglewood oil field operated more than 600 wells, and some of these wells were located as close as 200 meters from the dam. The results of investigation performed after the failure indicate that the failure was due to the breakdown of the underlying drainage system, which in turn, caused release of reservoir water that undermined the dam integrity. Several fault traces that are part of the Inglewood Fault System have also been mapped across the reservoir floor.
It has been speculated that the withdrawal and injection (flooding) activities on the nearby Inglewood oil field caused the area to subside and fault traces to move. These subsidence and fault movements (creeps) are believed to cause the failure of the drainage system and ultimately the dam itself.  Historically, the Guy and Greenbrier areas had experienced seismic activities (seismic swarms), which are believed to be associated with the nearby New Madrid Seismic Zone (located on the northeastern corner of Arkansas). The New Madrid Seismic Zone is the largest known seismic source east of the Rocky Mountains, and the source of the 1811-1812 great earthquake series (Mw 7.0 to 8.1) in the area.
The exact causes of the observed seismicity are not completely known. Judging from the seismicity pattern, however, it is conceivable that the recorded seismic events have direct correlation with the ongoing injection activities. It has been postulated that injection at Well #5, which intersects the Enders Fault, may allow the wastewater to migrate through the fault's planes into deeper crustal structures [17].
The pore pressure generated during injection may also reduce the contact stresses on the planes, causing the fault to slip in an earthquake.

Key Factors for DWI Induced Seismicity
There are about 151,000 Class II injection wells currently operating in the United States. Very few felt seismic events (i.e., events with Mw ≥ ±3.0 to 4.0) have been reported or documented as directly caused by wastewater disposal operations; the majority of these events were small magnitude earthquakes (i.e., events with Mw < 3.0 to 4.0) [2].
Accurate information on fluid injection and/or extraction is critical for assessing the potential of induced seismicity. The factors believed to be responsible for inducing seismic events are complex and interrelated. DWI induced seismicity is likely due to changes in insitu stresses in the Earth's crust caused by injected fluid pressure. The injected fluid pressure will reduce normal or contact stresses acting on a fracture's planes, which in turn, will reduce shear resistance to sliding. If reduction in shear resistance is large enough to cause slippage of the blocks, then an earthquake will occur. High fluid pressure can also change in-situ stress conditions within the solid rock formation and induce earthquakes.
The following key factors should be considered in assessing the potential of DWI induced seismicity [2]: • Earthquake history of the DWI area. Historical seismicity of the region provides the background or natural seismic environment prior to injection activities. It can be used as a basis to assess if any increased seismicity occurs naturally or is caused by injection activities.
High seismic activity, especially in tectonic plate boundary areas, may also indicate that the in-situ stresses on the Earth's crust are already in a delicate equilibrium state, and any disturbances from injection may induce earthquakes. • Presence of nearby fault(s). Injecting and/or extracting fluid near a known fault may alter the stress conditions on and along the fault's planes. Increased fluid pressure due to injection can reduce the contact stresses between plates, which in turn, will reduce their sliding resistance and lead to earthquakes. The vulnerability to produce fault slippage depends on fault activity, dimension, and orientation as well as on existing stress state. These include fault slip rate, strike, dip, and rake angles; top and bottom fault depths; and seismogenic depth and length. Distance of fault planes to injection well is also a critical factor in assessing the potential of inducing earthquakes. • Injection and/or extraction parameters. The parameters include rate of injection and/or extraction, duration of injection and/or extraction, volume and temperature of injected and/or extracted fluids, spatial distribution of wells, and generated pore pressure at depth. Observations made on some of the case histories have indicated that net fluid balance (i.e., total balance of fluid injected and extracted) appears to have the most impacts on pore pressure changes in the subsurface rock/soil over time; operation with balanced fluid volume seems to produce fewer induced seismic events. Reducing injection volume, rates, and pressure has also been successful in decreasing rates of induced seismicity. • Existing stress conditions. As discussed previously, injection pressure alters the in-situ stress equilibrium both in terms of stress amplitudes and principal directions in the Earth's crust. If the changes are significant enough, movements of existing fractures (or faults) could be initiated, leading to seismic events. The orientation of fracture planes with respect to principal stress directions determines the likelihood of generating seismic events. Faults with low-dipping angles (i.e., almost horizontal faults) should be less susceptible to stress-induced sliding. • Characteristics of target geologic strata. The characteristics of geologic strata that fluid is being injected into are important factors in the generation of induced seismic events. In fractured rocks, the injected fluid will travel along the network of fractural planes, and the impacts would largely be the reduction in contact stresses on the planes; whereas in permeable rocks, the fluid will migrate through the rock's pores and change the principal stress amplitudes and directions.

Conclusions
Several case histories related to increased or induced seismic events due to DWI were reviewed. Although cause-and-effect relationship of these induced seismic events is not exactly known, the case histories presented in this paper clearly indicate probable correlations between DWI activities and induced seismicity. The great majority of injection operations have not resulted in felt seismicity due to their low magnitudes, although earthquakes with Mw as large as 6.5 to 7 were detected beneath or near some of the oil and gas fields.