Geophysical data can provide some of the most spatially extensive information about the subsurface, and has a long and successful exploration role in the oil and gas industry [1
]. Seismic data, which utilizes elastic waves and how they reflect and refract off of materials with contrasting density and elastic moduli, has provided incredibly detailed models of faults, salt diapers and other important geologic features at depths reaching 100 km. Analogous to medical imaging technologies such as Ultrasound or magnetic resonance imaging (MRI) scans, which probe the structure of our body using physical fields, geophysical sensors (such as geophones or DAS sensors), are often placed at the surface, and measure physical fields, specifically in this paper, elastic waves. Like the ultrasound which probes the body with ultrasonic waves, active seismic sources such as dynamite explosions, vibroseis trucks or hammer swings, induce seismic waves probing the subsurface at increasing depths. Figure 1
depicts an incident primary wave, which is induced by a vibroseis source (the truck on the ground surface), and two possible reflected waves: P (primary-compressional) and S (shear) waves. The reflections occur because of changes in material properties such as density (ρ) and the acoustic velocities (Vp and Vs), according to Snell’s law [3
Geologic faults can provide abrupt changes in material properties, thus generating reflections and diffractions that can be recorded by seismological sensors. For geothermal applications, seismic methods have been used to locate faults, which will in turn help optimize the placement of future production wells, as faults can act as conduits for hot fluids to reach accessible depths [4
]. Melosh et al. [5
] describe the challenges of coherent noise masquerading as fault signatures in reflection seismic studies at the Blue Mountain geothermal field. However, recently, Huang et al. [6
] describe a state of the art anisotropic 2D migration technique applied to Blue Mountain, which revealed previously unseen faults from the traditional Kirchhoff approach.
Faults provide one of the principal means for a recharge of the geothermal reservoirs, particularly in the Basin and Range Region of Nevada, USA [7
]. The faults of Brady Hot Springs (located near Fernley, NV, USA) have been studied extensively [8
], as it is interpreted that they have allowed the near-continuous power production since 1991 via the recharge of water from the surface or re-injection [11
]. Previous analysis from InSAR and pumping data have hypothesized that the faults act as highly permeable conduits, which channel fluids from shallow aquifers to the deep geothermal reservoir tapped by the production wells [12
In particular, Siler and Faulds [10
] provided a 3D numerical fault model for the Brady system by integrating legacy 2D seismic reflection data, downhole lithologic data, and geologic map data. As stated in [10
], Brady is controlled by a step-over normal fault system, and the area contains active fumaroles, sinter, calcium carbonate tufa and silicified sediments. The authors refer readers to [10
] for an in-depth geologic review. Some of the faults are interpreted to have up to 10 m of displacement, and could thus provide reflections [13
]. Figure 2
displays the four major faults from the Siler & Faulds model. A study by Queen et al. [13
] concluded that vertical seismic profiling would be more successful than surface sensors at imaging the nearly vertical faults expected at Brady, and depicted in the Siler & Faulds model. Our results confirm the findings of the Queen et al. work.
In March 2016, a four-week experiment known as PoroTomo deployed an integrated technology in a 1500-by-500-by-400-m volume at Brady Hot Springs [14
]. The sensors and sources that are of interest for this paper are shown in Figure 2
. The 191 active vibroseis locations (known as “vibe points”) are shown as green dots, and the position of Well 56-1A, which contains the vertical fiber-optic cable for the Distributed Acoustic Sensing (DAS), is in blue. The DAS fiber is located from the surface to a depth of approximately 400 m.
The local PoroTomo coordinate system, seen in Figure 2
, is chosen such that the Y
axis is oriented along the major fault strikes, and the origin of this local system is close to the location of Well 56-1A. Note the three labeled faults in Figure 2
that we will image and compare using the synthetic and DAS field data in the Results Section.
DAS is a novel new type of seismic data, thanks to a combination of fiber-optic cable and advances in optical, light-pulse technology [15
]. Traditionally, seismic data is recorded by geophones or seismometers, which are coupled to the ground through spikes, and measure particle acceleration via voltmeters. Instead, DAS uses Rayleigh scattering in a fiber-optic cable to detect changes in strain rate due to seismic waves impinging on the fiber [16
]. DAS is most sensitive to waves that have particle motion parallel to the orientation of the fiber; DAS is insensitive to particle motion perpendicular to the fiber orientation (referred to as broadside insensitivity). Figure 1
nominally depicts how a vertically-deployed DAS system may be more sensitive to certain reflected waves due to a horizontal reflector. In this simplified case, at short offsets (smaller distances between the source and the vertical DAS, shown on the left side) the vertical DAS will be more sensitive to reflected P-waves rather than S-waves (or specifically SV-wave: shear-waves with particle motion in the vertical direction) because the particle direction (orange arrows) is inline with the fiber. DAS sensitivity may switch to be higher for reflected S-waves at larger offsets, loosely depicted on the right side of Figure 1
Recently, distributed acoustic sensing (DAS) has been used in the oil and gas industry along with and in place of geophones, as it provides denser spatial sampling than geophones [18
]. Countering the increased spatial resolution, however, is DAS’s lower signal-to-noise ratio, due to imperfect fiber-subsurface coupling and the directionality sensitivity, with the highest sensitivity to particle motion parallel to the fiber direction. The coupling of the fiber downhole can vary depending upon whether the fiber is installed behind casing or in tubing [20
]. This will directly affect the quality of signal measured by the fiber, and whether or not it is representative of the response of the surrounding geologic material.
The DAS research has largely focused on borehole environments in the oil and gas industry, because many wells are already equipped with fiber for monitoring production via other tools; therefore, vertical seismic profiles recorded by DAS are becoming more common place [21
]. Many studies focus on how the DAS measurements compare to raw observations with geophones [19
]. In the borehole environment, as depicted in Figure 1
, with approximately horizontal reflectors and relatively close distances between the vibroseis source and the well instrumented with DAS, reflected P-waves will provide particle motion in the vertical direction, thus parallel to the fiber. This will ensure recorded seismic DAS data that contains information about subsurface structure.
Three main challenges exist for the Brady DAS data. First, the fiber itself is not attached to the Well 56-1A infrastructure. This would have required a large investment in the well, such as removing or replacing the existing well casing. Second, the faults of Brady are not horizontal (Figure 2
), therefore the reflected waves will result in more complicated and varied particle directions. Lastly, the location of the Well 56-1A is in the extreme southwestern corner of the PoroTomo grid. The quality of the DAS signal will depend on the location of the vibe point, the angle of incidence with the fault, and the type of reflected wave (P- or S-wave). A more ideal acquisition geometry would have the DAS fiber centered among the source locations (e.g., vibe points), allowing for more recordings of the reflected energy. Permitting issues on historical sites and steam sinters were among the top considerations for how the PoroTomo survey was designed [11
Despite these challenges, the DAS data recorded a sufficient signal that allowed for 3D spatial models to be reconstructed. This paper will describe advanced imaging methods, not historically utilized for geothermal exploration, but popular in the oil and gas industry, that convert the DAS data into 3D reflectivity models that are consistent with the Siler & Faulds fault model. Specifically the reverse-time migration (RTM) method is described and applied to both synthetic and observed DAS data from Brady [23
]. RTM simulates the propagation of wavefields in 3D space, which is an improvement over traditional seismic modeling techniques that simplify the wave propagation via straight ray paths (as depicted in Figure 1
Two novel advances for geothermal exploration are presented in this paper. The first contribution is the application and verification of the DAS data to record useful seismic information, despite the challenging scenario at Brady. Second, this paper represents one of the first imaging experiments performed on DAS data, and the first for geothermal applications. The paper will first describe the DAS data and the method of reverse time migration (RTM). The next section will describe the results, which compare imaging using three different receiver wavefields: (1) Synthetic, (2) the DAS data from Brady and (3) pure noise. By comparing these results, we conclude which faults are likely “seen” by the DAS data, and we discuss the significance and possible improvements of this work.
Velocity uncertainty is one of the factors that reduces the ability of RTM to accurately image the location of the faults. The estimates of P-wave velocity versus depth around the location of Well 56-1A varies between different tomography methods [28
] and the interval velocity obtained directly from the DAS data [26
]. This, along with an imperfect fault model, that is based on legacy seismic and well information, will create uncertainty and variability within our images.
The effect of the velocity model can be seen in some of the RTM images. In the depth slice (upper left) of the a and b images in Figure 8
, the focusing effect of the velocity model is seen along the middle West-dipping fault (see Figure 4
for the general patterns of the velocity). Focusing is caused by patterns of higher velocities, in this case, these may be artifacts of the acquisition geometry itself [28
]. Focusing is also seen in the PoroTomoX and PoroTomoY panels. In the lower right panels, there are brighter spots that begin PoroTomoX = 500 and depth = 500, which coincide with a higher (4500 m/s versus 4000 m/s) velocity structure. There are also focusing patterns below the yellow box in the lower left panels.
Another modeling approach that could be considered is a Kirchhoff migration [30
]. Kirchhoff is ray-based, and although it may not capture the mode-conversion and scattering of wave-equation-based methods, it could result in fewer migration artifacts, which are highly present, given the non-ideal geometry of the vibe points, and the fiber in Well 56-A1. A more thorough analysis of the wavefields used during the imaging process with RTM may provide some angles of reflection, thus insight into the mode conversions by confirming the angles of incidence for the shot sources and faults [31
]. With angles of incidence, we may be able to explain where P to S mode conversions occur via Zoeppritz equations.
Future research directions should focus on further processing the DAS data to isolate events of interest. Before migrating, the field data was filtered to remove downgoing events via simple Fourier analysis. The field data would be less noisy with expert processing applied, such as statics corrections. The geophone data was not useful to this end, as it suffered from spatial aliasing (average spacing of 80 m), and the short maximum offset of 1500 m in the DAS horizontal array did not allow adequate recording of moveout (e.g., for identification of hyperbolas) [32
]. Also, this paper only utilized the P-mode sources (when the vibroseis truck “shook” up and down). Both shear-vertical (SV) and shear-horizontal (SH) sources were induced at every vibe point. Utilizing these multi-component sources would surely increase the information and certainty of the fault locations, as they may provide reflected energy that has particle direction aligned with the vertical fiber.
Because of Well 56-1A’s relative position to all the vibe points, it is very difficult to resolve any faults in the PoroTomoY plane that are close to or west of well location (PoroTomoX < 125 m). The results could be an argument for compressive sensing to help design acquisition geometry, especially when dealing with a limited budget and the restrictions such as fumaroles and permitting, which PoroTomo was with the Emigrant Trail [11
]. Compressive sensing exploits the fact that a small and carefully selected set of measurements of a compressible signal carries enough information for reconstruction and processing [33
As mentioned in the Introduction, DAS is considered to have a much lower signal-to-noise ratio compared to seismic data recorded on geophones. One issue is the gauge length, which is a type of spatial averaging done in the interrogator [15
], but another issue is also the coupling mechanism of the fiber to the subsurface. It is remarkable, but not unheard of, that the unattached, hanging DAS fiber is able to recover usable signal. Lindsey et al. [34
] describe how a variety of earthquakes are detectable using fiber deployed in standard telecommunications conduits, which are imperfectly coupled to the subsurface. Frictional coupling can allow for the fiber to record signal that is related to the subsurface media [35