Title: What’s Shaking in Bedrock? The Paradox Valley Deep-Well Injection Program
Authors: Jon Ake, Ken Mahrer, Lisa Block and Dan O’Connell, Seismotectonics and Geophysics Group, U.S. Bureau of Reclamation, Denver CO
Publication: The Outcrop, March 2002, p. 1, 12-14
“The Problem in Paradox”
The Colorado River provides water for nearly 23 million people and irrigation for about 1.6 million ha in the southwestern United States. However, excess salinity in the lower Colorado River results in economic damage and difficult political issues. Saline-saturated ground water discharges into the Dolores River, a tributary of the Colorado, as it crosses the Paradox Valley in southwestern Colorado, increasing the dissolved-solids load by approximately 200,000 tons per year. Because this is one of the most significant point sources of salt into the Colorado River system, it prompted construction of the Paradox Valley Unit (PVU), one of numerous salinity control projects developed by the Bureau of Reclamation within the Colorado River basin. The estimated economic benefit of salinity control is $340/ton of salt removed (1994 dollars). Approximately 650,000 tons of salt have been removed by the PVU since 1991.
At present, PVU consists of 13 shallow (12-21 m depth) brine extraction wells along the Dolores River, a surface treatment facility, a deep (4.9 km) injection well and a local seismic monitoring network. Prior to injection the brine is cut with 30% fresh water to prevent in situ calcium sulfate precipitation and treated with a corrosion inhibitor. The injection horizon is the Mississippian-age Leadville Limestone at a depth of ~4.3 km. The surface facilities (which operate 24 hrs per day) are capable of injecting 1325 liters/min at a sustained pressure of 34 MPa. Since operations began in 1991 , the PVU has injected approximately 3 billion liters of fluid.
The town of Bedrock is located within the Paradox Valley about 120 km south of Grand Junction, CO (Figure 1). The valley is a collapsed diapiric salt anticline that trends approximately N55°W; is 38 km long, 5-7 km wide; and is underlain by up to 6 km of interbedded salts and shales of the Pennsylvanian Paradox Formation. The Dolores River crosses Paradox Valley (normal to strike and across the anticline, hence the “paradox”) after flowing through deeply incised canyons. Gently dipping late Jurassic and Triassic sedimentary rocks form spectacular mesas and canyon walls at the surface. The target Leadville Limestone is a locally vuggy, fractured dolomitic limestone with an effective porosity less than ~6%. However, widespread fracturing significantly increases the local permeability. Petroleum industry investigations and PVU design studies identified a series of Laramide-age faults (the Wray Mesa fault system) trending sub-parallel to the strike of the Paradox Valley. The injection well was sited to optimize fluid migration along these pre-existing fractures.
Testing at PVU began in 1991. Seven injection tests, with durations ranging from a few weeks to months, were conducted between July 1991 and March 1995. During these tests injection rate and chemistry of the injectate were varied. Continuous injection began in July 1996. Pressure and injection rate are measured at the surface; no downhole measurements are available during injection.
Paradox Valley Seismic Network
The potential to induce earthquakes by high-pressure, deep-well injection was conclusively established at the Rocky Mountain Arsenal near Denver in the 1960’s. Recognizing this potential, the Paradox Valley Seismic Network (PVSN) was installed in 1984. The network presently operates 15 stations loosely arranged in two concentric rings about the injection well; the outer ring has a diameter of about 80 km. All sites have extremely low levels of cultural noise (evident to anyone visiting Bedrock and the Paradox Valley), which allows detection and location of very small earthquakes. The seismic monitoring program: (1) records, evaluates and locates seismic events in the region surrounding the Paradox Valley, with a specific focus on the immediate vicinity of the injection well; (2) determines focal mechanisms of the events when feasible; and (3) identifies and evaluates relationships between seismicity, geology, subsurface brine location, and injection parameters.
In almost six years of monitoring prior to the first injection test in 1991, PVSN recorded and located only six earthquakes within the 19,000 km2 region monitored by the array. None of these events were within 5 km of the injection well. During the first week of the initial injection test in 1991, more than a dozen earthquakes were detected and located within ~1 km of the well. Between July 1991 and March 2001, PVSN has recorded and located more than 3,700 events within 10 km of the injection well. About a dozen events were large enough to be felt (magnitude (M) 2.5 or greater). The largest event, an M 4.3 event, occurred on May 27, 2000. Based on the microearthquake locations, we infer that fluid-pressure perturbations have migrated at least 8 km from the injection well. A strong association between injection and seismicity has been observed throughout the course of the project. In mid-1999, an M 3.7 earthquake caused PVU operations to include a 20-day injection shutdown every six months. Prior to the May 27th M 4.3 earthquake, PVU injected fluid at ~1290 liters/min with a nominal surface pressure of 33.1 MPa. Following the May 27th event, well operations were suspended for 28 days. To reduce the potential for more large events we resumed pumping on June 23rd at a reduced rate of ~870 liters/min. The 20-day shutdown periods and lower injection rate have substantially reduced the rate of earthquake inducement as shown in Figure 2.
To enhance our understanding of injectate and connate fluid flow, we have focused on improving earthquake location procedures. We performed a three-dimensional analysis of approximately 650 events and obtained a new three-dimensional P-wave velocity model. The model reproduces the major geological features including the low-velocity region of the Paradox salt anticline and high-velocity region of the laccolith beneath the LaSalle Mountains to the north. Using this velocity model and seismic wave travel-time differences we improved the mapped locations of about 3500 events. We found that the earthquakes were confined to a relatively narrow depth range, with most dipping gently to the northeast following the inferred depth interval of the Leadville Formation (see Figure 3). The grouping of events approximately 0.5 km to the southwest of the injection well agrees with the inferred location and trend of one of the faults of the Wray Mesa fault system. A second distinct group of events 2.5 km southwest of the well is interpreted to be a previously unidentified element of the same system. The deepest events near the well occur below the Leadville within the crystalline Precambrian rocks.
Using 346 events, we calculated seismic focal mechanisms. The results suggest mainly strike-slip motion with two dominant fault plane orientations. The direction of minimum stress (T-axis) is consistently northeast and sub-horizontal for all events. Analysis of pressure data from well completion showed relatively high deviatoric stresses (s1-s3). These observations suggest the maximum and intermediate principal stresses are nearly equal at the injection depth and considerably larger than the minimum principal stress. Fractures observed in oriented core samples recovered during drilling agree with the strike of the observed focal planes. The event locations define lineaments having strikes consistent with the fault planes suggested by the focal mechanisms. The lack of microearthquake fault planes with orientations parallel to the major throughgoing faults of the Wray Mesa system suggests these N55°W striking planes may be conduits for fluid transport but do not have sufficient shear stress to produce earthquakes. They are favorably oriented for dilation (normal to the northeast-directed minimum principal stress) within the inferred stress field.
By understanding the relationship between small earthquakes and injection parameters (in particular injection rate) we have been able to modify operations at PVU to minimize the likelihood of future larger, damaging earthquakes. We are pleased to report the classic Bedrock general store is undamaged.
These studies were possible through the continued support of Andy Nicholas, Project Manager of the Paradox Valley Unit. Mike Sullivan contributed Figure 3; Donna and Larry Anderson provided helpful review comments and the great picture of the Bedrock store.