InSAR Measurements and Numerical Models of …...X!!!−X!!! where Δ! is the range change, and X the easting coordinate, both of which have dimensions of length. We choose the east

PROCEEDINGS, Thirty-Ninth Workshop on Geothermal Reservoir Engineering Stanford University, Stanford, California, February 24-26, 2014 SGP-TR-202

1

InSAR measurements and numerical models of deformation at Brady Hot Springs geothermal field (Nevada), 1997-2013

S.T. Ali1, N.C. Davatzes2, P.S. Drakos3, K.L. Feigl1, W. Foxall4, C.W. Kreemer5, R.J. Mellors6, H.F. Wang1 and E. Zemach3

[1] Department of Geoscience, University of Wisconsin, Madison WI, USA

[2] Department of Earth and Environmental Science, Temple University, Philadelphia PA, USA

[3] Ormat Technologies Inc., Reno NV, USA

[4] Lawrence Berkeley National Laboratory, Berkeley CA, USA

[5] Nevada Bureau of Mines and Geology, University of Nevada, Reno NV, USA

[6] Lawrence Livermore National Laboratory, Livermore CA, USA

Email: [email protected]

Keywords: EGS, InSAR

ABSTRACT InSAR images acquired by the ERS and TerraSAR-X satellites between 1997 and 2013 over the Brady Hot Springs geothermal field delineate subsidence on the order of a few centimeters per year over an elliptically shaped area roughly 5 km long by 2 km wide. This subsiding area is centered adjacent to a prominent bend in the fault system where the successful production wells are located. The long axis of the deforming region parallels the north-northeast strike of the predominant normal fault system. Within this broad bowl of subsidence, the interference pattern shows several smaller features with length scales of the order of ~1 km. Inverse modeling suggests that these small scale features are a result of contraction, following pressure decline, in shallow laterally confined reservoirs. Using poroelastic models we demonstrate how highly permeable conduits, associated with faults, can channel fluids from the shallow reservoirs to the deep reservoir which is tapped by the production wells. Such structurally controlled, high permeability conduits are consistent with relatively recent fault slip evidenced by scarps in late Pleistocene Lake Lahontan sediments and spatially associated surface hydrothermal features that predate production at Brady. In contrast, Desert Peak, a “blind” geothermal field, located less than 7 km away from Brady, shows little or no deformation in the InSAR dataset, although the two fields are otherwise similar in spatial extent, structural setting, and geothermal production. Desert Peak exhibits neither hydrothermal features nor any evidence of recent surficial fault slip, however, suggesting that the “plumbing” associated with the fault system there is deeper and more isolated from the surface than at Brady.

1. INTRODUCTION The Brady Hot Springs Geothermal field is located about ~80 km east-northeast of Reno, in the Hot Spring Mountains of northwestern Nevada. The area surrounding the field is dominated by a network of north-northeast trending, steeply dipping, en echelon normal faults, as shown in Figure 1. A ~20 megawatt enhanced geothermal power plant at Brady has been generating power since 1992. Six production wells, located near a prominent bend in the normal fault system (Figure 1) are used to withdraw hot water from depths of 0.5-2 km. Following generation of electricity, the water is recycled back into the subsurface, between depths of 0.5-1 km, via two injectors located 1.5-2.5 km away in the north-northeast direction, along the strike of the predominant fault system. The reservoir is hosted in layered Tertiary volcanic rocks including welded tuff, rhyolite and meta-sediments overlying Mesozoic crystalline intrusions. The primary production interval reaches temperatures of 175-205 C at a relatively shallow (1-2 km) depth (Benoit and Butler, 1983; Shevenell et al., 2012). Figure 2 shows the pumping history at Brady during the last decade. The average, net extraction rate over the time interval is ~0.1 m3/sec.

Withdrawal of fluids from the subsurface changes the pore pressure field inside the reservoir which results in surface deformation that can be measured using interferometric synthetic aperture radar (InSAR). InSAR uses the phase difference between two SAR images, acquired over two different time periods, to create an interferogram (Massonnet and Feigl, 1998). The resulting ambiguous wrapped phase contains information about ground deformation, topography and atmospheric noise. After correcting the interferogram for topography the wrapped phase is unwrapped, using an unwrapping algorithm, to estimate the range change towards the satellite. By analyzing the pattern of deformation we can gain insight into the underlying geomechanical processes and “plumbing” of the reservoir which can aid the plant operator to better manage the resource. In recent years, InSAR has been used to study deformation at several geothermal fields,

Ali et al.

2

including, Cerro Prieto (Carnec and Fabriol, 1999; Sarychikhina et al., 2007), Coso (Fialko and Simons, 2000; Wicks et al., 2001), Dixie Valley (Foxall and Vasco, 2003), Brady (Oppliger et al., 2004, 2006; Shevenell et al., 2012), Taupo Volcanic Zone (Chang et al., 2005; Hole et al., 2007), Svartsengi (Jonsson, 2009; Masters, 2011), Imperial Valley (Eneva et al., 2009, 2012), San Emidio (Eneva et al., 2011), Geysers (Vasco et al., 2013) etc. In these studies, single pairs or multi-baseline interferogram stacks were used to map cumulative deformation following geothermal production. Oppliger et al. (2004, 2006) were the first to apply InSAR at Brady Hot Springs. They mapped the cumulative deformation resulting from geothermal production between 1997 and 2002, using images acquired by the C-band ERS satellite of the European Space Agency (ESA), and interpreted the signal in terms of a contracting aquifer.

Figure 1 Map showing location of the Brady Hot Springs geothermal field. The fault map is from Faulds et al. (2010) and surface hydrothermal activity is from Coolbaugh et al. (2004). Injectors are shown by blue triangles and producers are shown by red triangles. The stress model, from Moos (unpublished report, 2011), is summarized as an outer red ring centered on 15-12 corresponding to the vertical stress. The radial length of the other stress components are scaled to the vertical stress; the inner blue ring corresponds to the static fluid pressure; black lines and yellow wedges correspond to the mean horizontal principal stress directions ±1 standard deviation. The interferogram in the background shows wrapped phase change values for pair 13 (Table 1), spanning the 308-day interval from Dec-24-2011 to Oct-27-2012. One colored fringe corresponds to one cycle of phase change, or 16 mm of range change.

Figure 2 Pumping history at Brady Hot Springs, showing the volumetric rates (upper panel) and cumulative volume (lower panel) for injection (blue curves), production (red curves), and net injection (black) as measured at well heads (data courtesy Ormat).

Ali et al.

3

In this study we use multiple images acquired between 2011 and 2013 by the X-band TerraSAR-X (TSX) and TanDEM-X (TDX) satellites of the German Space Agency (DLR) along with older archived ERS images, acquired between 1997 and 2004, to constrain geomechanical models in order to characterize time dependent deformation at Brady. The high quality, high-resolution, short-term TSX/TDX interferograms show multiple small length-scale features, aligned along the direction of the strike of the of the predominant normal fault system, that we attribute to contraction of shallow reservoirs following dewatering. Using a model, based on the theory of poroelasticity (Wang, 2000), we demonstrate how faults or damaged zones near the faults can act as conduits, channeling fluids from multiple, laterally confined, shallow reservoirs to the deeper reservoir tapped by production wells, resulting in localized deformation.

2. DATA AND METHODS 2.1 SAR Interferometry We analyze 28 SAR images acquired by the ERS and TSX/TDX satellites between 1997 and 2013 and combine them to form 29 distinct interferometric pairs as listed in Table 1. We generate the interferograms using the DIAPASON software developed by the French Space Agency CNES, and the DTOOLS post-processing package of scripts (Massonnet and Rabaute, 1993; Massonnet and Feigl, 1998). The wrapped phase values are filtered using their two dimensional spectra (Goldstein and Werner, 1998). The topographic contribution to the interferograms is removed using a 10 m National Elevation Dataset digital elevation model (DEM) (Gesch et al., 2002). For example, the observed wrapped phase change values for the TSX pair spanning the 308-day time interval, between 2011.9781-2012.8197, is shown in Figure 1. One fringe of phase change in this interferogram corresponds to 16 mm of range change along the radar line of sight between the satellite and the ground.

Figure 3 Interferograms showing observed values of wrapped phase change for all TSX/TDX pairs in Track 167, listed in Table 1, sorted by the number of days spanned. One colored fringe corresponds to one cycle of phase change, or 16 mm of range change

The interferogram shows multiple signatures of varying length-scales. For example, a broad elliptically shaped area roughly 5 km long by 2 km wide, outlined by the dashed black line, is observed in the area, consistent with earlier observations (Oppliger et al., 2004, 2006). Within this broad bowl we observe several smaller elliptic/circular features or fringe patterns, with length scales ranging from 1-2 km, both near the production wells as well as near the two injectors as shown in Figure 1. The features are aligned parallel to the trend of predominant fault system, including the 4-km-long Brady normal fault, and associated fumaroles. This multiscale signature is consistently observed in all interferograms spanning more than a year, e.g., in all 11 TSX/TDX pairs in Track 167 as shown in Figure 3. Pairs spanning a shorter time interval, i.e., less than year, only show the small length-scale features near the producers and the injectors. Using pairwise logic, we reject the possibility that the observed signatures are due to imprecise orbital trajectories or atmospheric

Ali et al.

4

perturbations. Since the shape of the fringe pattern does not resemble the topographic relief, we reject the possibility that the signature is due to errors in the DEM.

2.2 Inverse modeling In order to gain more insight into the sources behind these small length-scale features observed in the interferograms, their geometry, as well as location, we perform non-linear inverse modeling using the range gradient as the observable quantity. Following Ali and Feigl (2012), we calculate the dimensionless gradient of range change 𝜓 ! for the kth pixel as:

𝜓 ! =Δ𝜌 !!! − Δ𝜌 !!!

X !!! − X !!!

where Δ𝜌 is the range change, and X the easting coordinate, both of which have dimensions of length. We choose the east component of the range change gradient because it is sensitive to the details of the deformation. Just as the range change measured by InSAR is sensitive to displacement, the range change gradient is sensitive to strain. Mathematically, the range change gradient is equivalent to a particular component of the deformation gradient tensor (Malvern, 1969). Since the range change gradient is a continuous and differentiable quantity (Sandwell and Price, 1998), using it as the observable avoids the pitfalls associated with phase unwrapping techniques. The range change gradient values are derived from wrapped phase data by a quadtree resampling procedure, as summarized by Ali and Feigl (2012). The resampling reduces the computational burden and mitigates correlations between neighboring pixels. For example, the quadtree resampling procedure reads the phase values shown in Figure 1 to extract the range change gradient shown in Figure 4e. Each interferogram is reduced to 1000-5000 values of the range change gradient. The corresponding resampled phase values are shown in Figure 4a.

Figure 4 (a) Interferogram showing observed values of wrapped phase change for pair 13 (Table 1), spanning the 308-day interval from Dec-24-2011 to Oct-27-2012. One colored fringe corresponds to one cycle of phase change, or 16 mm of range change. (b) Modeled wrapped phase values calculated from the final estimate of the parameters in the elastic model. (c) Observed values of the range gradient in cycles per pixel after resampling by the quadtree algorithm. (d) Modeled values of the range change gradient calculated from the final estimate of the parameters in the elastic model, following inversion.

In order to describe the signatures observed in the interferograms, we use a purely elastic geomechanical model where all deformation is attributed to changes in volume of the reservoirs. We model the deformable reservoirs using a combination of point sources (Mogi, 1958), also known as Mogi sources, and mode 1, rectangular tensile dislocations (Okada, 1985), or Okada sources, in an elastic halfspace with uniform material properties. The Okada sources are especially appropriate because the thickness of a reservoir is usually much smaller than its length and width. By estimating the parameters in this “idealized” geomechanical model we can gain insight into the geometry and location of the reservoir, e.g., its extent and depth, as well as the associated volume change consistent with the surface deformation, and presumably due to injection/extraction. Because the computational cost associated with these semi-analytical halfspace models is minimal, we use the global optimization method of simulated annealing (Kirkpatrick et al., 1983) to estimate the parameters. We perform the inversion using the “general inversion for phase technique” (GIPhT), developed by Feigl and Thurber (2009) and extended by Ali and Feigl (2012). GIPhT minimizes the misfit between the observed and modeled values of the range gradient, averaged over all points in the resampled dataset. To calculate the uncertainty of estimated parameters, we set the 68-percent critical value of the cost to be the final cost plus a quarter of the difference between its initial and final value.

Ali et al.

5

Since this approach yields unrealistically small values, we re-scale the uncertainties by the square root of the mean-squared error to find the 1σ uncertainties.

Figure 5 Subsidence rate calculated from the final estimate of the parameters in the elastic model for pair 13 (Table 1) shown in Figure 4. Producers and injectors are shown by triangles and inverted triangles, respectively.

Our geomechanical model includes two Mogi and two Okada sources. The free parameters for each Mogi source includes its position and volume change. For each Okada source the free parameters include, position, geometry (i.e., length, width, dip and strike) and the opening. We first perform inversion using the resampled data for each pair individually. For example, Figure 4 shows the result for pair number 13 (Table 1), spanning the 2011.9781-2012.8197 time interval. Figure 4a shows phase values after quadtree resampling. Similarly, Figure 4c shows the observed values of the range change gradient estimated during quadtree resampling. The modeled values of the range change gradient calculated using the final estimate are shown in Figure 4d. During the inversion the cost, defined as the misfit of the range gradient averaged over all pixels, decreases from 0.006 cycles per pixel to 0.005 cycles per pixel. The main features, such as the two (yellow and blue) large elliptical lobes near the production wells in the south, and the two circular lobes near the injectors, are reproduced by the model. Similarly, the modeled values of phase (Figure 4b), calculated using the final estimate of the parameters, are in good agreement with the observed values (Figure 4a).

Once we have the final estimate of the parameters we can also use it to calculate other quantities of interest, e.g., the subsidence rate, as shown in Figure 5 or the rate of volume change, due to all four sources, as plotted in Figure 6. We repeat this procedure for all other 28 pairs. We compare the results by plotting the estimated rate of volume change, in m3/year, for each pair as shown in Figure 6. All but two of the analyzed pairs show a net negative extraction rate at Brady Hot Springs. The two pairs that do show a net positive extraction rate, span a very short time interval and have a low signal-to-noise ratio. The net rate of volume change at Brady over 1997-2013, assuming an elastic halfspace model, averaged over all 29 pairs is 25E3 m3/year or about 6.5E-4 m3/sec. This value is less than 1-percent of the net extraction rate estimated by the operator from well data. Also, the depth estimate of all Mogi and Okada sources in our model, averaged over all pairs, is between 100-425 m (Figure 6), much shallower than the depth of production wells which averages ~1 km. We obtain a similar result when we perform inversion by combining data from multiple pairs into an ensemble dataset as shown by the red squares in Figure 6. This suggests that the inversion procedure is robust. The values of the estimated parameters are listed in Table 2. Figure 7 shows the average subsidence rate, calculated using parameters estimated from the ensemble dataset. The areas associated with subsidence are located right above the 4 km long Brady normal fault as shown in Figure 7 (top panel). A peak subsidence rate of ~20 mm/year is estimated for the area 0.5-1 km south/southeast of the producers. The subsidence rate near the injectors is estimated at ~15 mm/year.

To explain the result, we hypothesize that the deformation observed in the interferograms is caused by pressure decline in multiple, small contracting reservoirs in the shallow subsurface, i.e., between 0-500 m. These shallow, laterally confined reservoirs, potentially comprising of highly porous and compressible sediments, lose mass as fluids are channeled via permeable normal faults (e.g., the 4-km long, north-northeast-striking, west-northwest-dipping Brady’s fault and associated fault strands) or damaged fault zones, as shown in Figure 7, to the production wells. This phenomenon explains the contracting reservoirs near the injectors as injection is typically associated with uplift. It also explains the fumarolic activity in the area as colder fluids migrate from shallow reservoirs to the hotter, deeper reservoir. The reason we see a large discrepancy between the estimated (6.5E-4m3/sec) and measured (0.1 m3/sec) volume loss is because, either the main reservoir, which provides most of the hot water at Brady, is possibly deeper than 1 km, resulting in a much larger wavelength subsidence signal that is not resolvable in interferograms which we have analyzed, or perhaps because fluids

Ali et al.

6

are being sourced from areas not in the vicinity of the plant through the faults. Some interferometric pairs, spanning a much larger time interval (~5-6 years), that were analyzed by Oppliger et al. (2005) do indicate that sources of deformation at Brady are perhaps as deep at 1.5 km, consistent with successful production wells.

Figure 6 Blue crosses represent the parameter value for each of the 29 interferometric pairs, along with the uncertainty. Black square and green circle represents the unweighted and weighted mean respectively. The red square shows the parameter value estimated from the “ensemble” dataset.

Ali et al.

7

Figure 7 Subsidence rate, calculated using parameters estimated from the ensemble dataset. The perspective view of the subsurface shows various faults (cyan) including the Brady normal fault (blue).

2.3 Heuristic poroelastic modeling To illustrate our hypothesis of shallow dewatering via faults, which provide a path for fluid migration, and to explore its implications, we use a fully numerical poroelastic model that simulates deformation due to production and injection at Brady. We perform the simulation using Defmod (Ali, 2011, 2014), an open-source, fully unstructured, implicit/explicit, two or three dimensional, parallel multiphysics finite element code. Defmod was originally designed for modeling deformation due to earthquakes/volcanoes over time scales ranging from milliseconds to thousands of years and can model processes such as co-seismic slip or dike intrusion(s), poroelastic rebound due to fluid flow and post-seismic or post-rifting viscoelastic relaxation. However, it can also be used to model processes such as post-glacial rebound, hydrological (un)loading, injection and/or withdrawal of compressible or incompressible fluids from subsurface reservoirs etc. The code is written in Fortran 95 and uses PETSc’s (Portable, Extensible Toolkit for Scientific computation, Balay et al., 2012) parallel sparse data structures and implicit solvers. Problems can be solved using (stabilized) linear triangular, quadrilateral, tetrahedral or hexahedral elements on shared or distributed memory machines with hundreds or even thousands of processor cores. For the quasistatic poroelasticity problem Defmod solves the coupled system of equations, comprising of the equilibrium and the continuity equation which, following Zheng et al. (2003), is written in the semi-discretized form as:

𝐾!𝑢 − 𝐻𝑝 = 𝑓

𝐻!𝑢 + 𝑆𝑝 + 𝐾!𝑝 = 𝑞

where 𝑢 and 𝑝 are displacements and pressure, respectively, 𝐾! is the elastic stiffness matrix for the solid frame, 𝐻 is the coupling matrix representing the poroelastic expansion due to fluid pressure change, 𝑆 is the matrix representing change in fluid storage due to pore fluid pressure change, 𝐾! is the permeability matrix, and the right hand side represents boundary forces 𝑓, and fluid sources or sinks 𝑞. The time dependent system of equations is solved using an incremental loading scheme (Smith and Griffiths, 2005). To circumvent the Ladyzenskaja-Babuska-Brezzi restrictions on linear elements, Defmod uses the local pressure projection scheme proposed by Bochev and Dohrmann (2006). The scheme works well for linear quadrilateral and hexahedral elements (White and Borja, 2008). It works well for linear triangles and tetrahedrons as long as a higher order integration scheme is used, i.e., 3-point integration for triangles and 4-point integration for tetrahedral elements. By using stabilized linear elements Defmod can solve poroelasticity problems more efficiently than codes that use quadratic approximation for displacements and linear for pressure.

To perform the simulation, we use a three dimensional conceptual model with two laterally confined reservoirs, located between depths of 100-500 m. We discretize the geometry using linear hexahedral elements. The mesh spans 10 km in length (X-axis), 10 km in breadth (Y-axis), and 5 km in depth (Z-axis) and contains ~650K degrees of freedom. The

Ali et al.

8

domain is large enough to avoid edge effects. The length of a single element’s edge varies from ~500 m at the bottom boundary to ~50 m near the fault zone. We assume a constant Poisson’s ratio of 0.25 and a Young’s modulus of 30 GPa for the entire domain. For simplicity, we assume a constant permeability of 0.1 darcy for reservoirs as well as the faults, and 1.0E-9 darcy for the surrounding, low permeability rock. These values are similar to those used in previous modeling studies of geothermal systems in the Basin and Range (e.g., Fairley, 2009; Moulding and Brikowski, 2012). We assume a Biot coefficient of 0.9, a porosity of 0.1 and a fluid bulk modulus of 2.2GPa. We neglect the effect of conductive and convective heat transfer. The geometry of the fault has been chosen arbitrarily so as to connect the shallow reservoirs to the deep reservoir which is assumed to be 4 km long, 2 km wide and 0.5 km thick. It only implies “the connection(s)” that are needed to represent the complex fault system at Brady at the conceptual level in a three-dimensional model. Existence of such permeable faults in geothermal systems within the Great Basin is well documented (e.g., Caine et al., 1996; Fairley and Hinds, 2004; Faulds et al., 2011). A net withdrawal rate of 0.1 m3/sec is assumed, consistent with pumping records. The simulation is run for 10 years during which deformation occurs due to net extraction of fluids. We consider two end member cases. In the first case the fault has the same permeability as the reservoir. In the second case the fault is assumed to be impermeable and has the same permeability as the surrounding rock.

Figure 8 Pore pressure field due to production, with (top panel) and without (bottom panel) a highly permeable, idealized, vertical fault which connects the two shallow reservoirs to the deep reservoir where the wells are located. Note: Only a subset of the mesh is shown.

Figure 8 shows a subset of the deformed mesh along with the pore pressure field at the end of the simulation for both cases, i.e., with and without the highly permeable fault zone. In Figure 8b the entire reservoir contracts, causing the subsidence signature at the surface to have a very long wavelength which is unlikely to be detected by conventional InSAR. However, in Figure 8a, we observe multiple areas of localized subsidence, similar to those observed using InSAR at Brady, as fluids are channeled via the fault from the shallow reservoirs to the deeper reservoir which is tapped by the production wells. Such a scenario would also explain the mismatch between the InSAR estimated, and the measured net extraction rate, as the longer wavelength signature is too small to be seen in any of the 29 interferometric pairs, spanning a relatively short period (between 121-550 days), leading to an underestimation of the total volume loss. According to our working hypothesis, Figure 8a is an idealized representation of the enhanced geothermal system (EGS) at Brady whereas Figure 8b, potentially is representative of the EGS at Desert Peak, a blind geothermal field, located only 7 km away from

Ali et al.

9

Brady, where no deformation has been observed in the any of the 29 interferometric pairs used in this study (e.g., Figure 9). Desert Peak is very similar to Brady as far as the setting, placement of wells and geothermal production is concerned. We observe little deformation at Desert Peak because unlike Brady, the caprock there is not breached by permeable faults which could channel fluids between the potentially shallow and deep reservoirs. Desert Peak also does not show any evidence of fumarole activity or recent fault slip, unlike Brady (Caskey and Wesnousky, 2000).

Figure 9 Interferogram showing wrapped phase change values for pair 13 (Table 1) both at Brady and at Desert Peak. One colored fringe corresponds to one cycle of phase change, or 16 mm of range change.

In order to better constrain the geometry of the idealized “plumbing system” at Brady, as highlighted by the -50 MPa pore pressure contour in Figure 10, we are currently in the process of integrating the poroelastic models with GIPhT. The inverse modeling will allow us to estimate parameters in the poroelastic model that best fit the InSAR data. Because we have a computationally expensive forward problem, we will use a local, gradient based inversion scheme (e.g., Ali et al., 2014). We will parameterize the geometry of reservoirs using simple shapes such as ellipsoids and cuboids (e.g., see Figure 8a) within a semi-structured grid. Free parameters in the model will include the geometry and depth of the reservoirs. We will also incorporate geologic information Faulds et al. (2010) into the mesh using the digital geologic database developed by Jolie et al. (2012), that includes all major faults (e.g., Figure 7) as well as geologic horizons.

CONCLUSIONS In this paper we use InSAR data acquired by the ERS and TerraSAR-X/TanDEM-X satellites to study deformation caused by production at the Brady Hot Springs geothermal field. The high quality, high-resolution, short-term TerraSAR-X/TanDEM-X interferograms show multiple signatures of varying length-scales associated with subsidence. Inverse modeling, using halfspace elastic models, suggests that the sources causing the deformation are located between depths of 100-425 m, which is much shallower than the average depth of production wells. Such a phenomena has previously been observed at the Dixie Valley geothermal field where deformation has been attributed to shallow sources within the valley fill above the reservoir (Foxall and Vasco, 2003). Based on the mismatch between the estimated and measured volume change at Brady, and the presence of multiple, locally subsiding regions, specially near the injectors, we hypothesize that a complex network of permeable faults (e.g., the 4-km long Brady’s fault which potentially has a high level of transmissivity) is channeling fluids from shallow, laterally confined, reservoirs to the deep reservoir where the production wells are located. The contraction of the deep reservoir likely produces a large wavelength signal that cannot be easily seen in the 29 pairs analyzed in this study. Such a scenario would also explain the lack of deformation observed at Desert Peak. Continued acquisition of radar data by the TerraSAR-X/TanDEM-X satellites in the near future will allow us to create longer time span interferograms which are key for observing and measuring the large wavelength subsidence signal.

Ali et al.

10

Such interferograms will allow us to constrain the geometry of the deep reservoir both at Brady and Desert Peak. Our study demonstrates that SAR interferometry is a useful tool which can provide valuable insight into the complex geomechanical processes within an EGS system.

Figure 10 The blue surface, on the right half of the clipped mesh, represents the -50 MPa pore pressure contour, after 10 years of production for the model shown in Figure 8a.

ACKNOWLEDGMENTS We thank (i) Ormat Technologies for collaboration and access to pumping records, (ii) Jeff Wagoner for providing the digitized three dimensional geologic model for Brady developed by Egbert Jolie, Inga Moek, and Jim Faulds (iii) Paul Spielman and John Akerley for their guidance and input with regard to the production history at Brady and Desert Peak. This project was supported by the U.S. Department of Energy’s Geothermal Technologies Office award DE-EE0005510. SAR data from the TerraSAR-X and TanDEM-X satellites operated by the German Space Agency (DLR) were used under the terms and conditions of Research Project RES1236.

REFERENCES Ali, S., Feigl, K., Carr, B., Masterlark, T., and Sigmundsson, F. (2014). Geodetic measurements and numerical models of

rifting in Northern Iceland for 1993-2008. Geophysical Journal International, 196(3):1267-1280.

Ali, S. T. (2011). Defmod - Finite element code for modeling crustal deformation. http://defmod.googlecode.com.

Ali, S. T. (2014). Defmod - Parallel multiphysics finite element code for modeling crustal deformation during the earthquake/rifting cycle. arXiv preprint arXiv:1402.0429.

Ali, S. T. and Feigl, K. L. (2012). A new strategy for estimating geophysical parameters from InSAR data: application to the Krafla central volcano in Iceland. Geochemistry Geophysics Geosystems, DOI: 10.1029/2012GC004112.

Balay, S., Brown, J., Buschelman, K., Gropp, W. D., Kaushik, D., Knepley, M. G., McInnes, L. C., Smith, B. F., and Zhang, H. (2012). PETSc (Portable, Extensible Toolkit for Scientific computation) web page. http://www.mcs.anl.gov/petsc.

Benoit, W. R. and Butler, R. W. (1983). A review of high-temperature geothermal developments in the northern Basin and Range province. Geothermal Resources Council, 1:57-80.

Bochev, P. B. and Dohrmann, C. R. (2006). A computational study of stabilized, low-order C finite element approximations of Darcy equations. Computational Mechanics, 38(4-5):323-333.

Caine, J. S., Evans, J. P., and Forster, C. B. (1996). Fault zone architecture and permeability structure. Geology, 24(11):1025-1028.

Carnec, C. and Fabriol, H. (1999). Monitoring and modeling land subsidence at the Cerro Prieto geothermal field, Baja California, Mexico, using SAR interferometry. Geophysical Research Letters, 26(9):1211-1214.

Ali et al.

11

Caskey, S. J. and Wesnousky, S. G. (2000). Active faulting and stress redistributions in Dixie Valley, Beowawe, and Bradys geothermal fields: Implications for geothermal exploration in the Basin and Range. In Proceedings. Twenty-fifth Workshop on Geothermal Reservoir Engineering. Stanford Geothermal Program Workshop Report SGP-TR-165. Stanford University.

Chang, H.-C., Ge, L., and Rizos, C. (2005). Insar and mathematical modelling for measuring surface deformation due to geothermal water extraction in New Zealand. In International Geoscience and Remote Sensing Symposium, volume 3, page 1587.

Coolbaugh, M. F., Sladek, C., Kratt, C., and Edmondo, G. (2004). Digital mapping of structurally controlled geothermal features with GPS units and pocket computers. Geothermal Resources Council Transactions, 28:321-325.

Eneva, M., Adams, D., Falorni, G., and Morgan, J. (2012). Surface deformation in Imperial Valley, CA, from satellite radar interferometry. Geoth. Resour. Counc.-Trans, 36:1339-1344.

Eneva, M., Falorni, G., Adams, D., Allievi, J., and Novali, F. (2009). Application of satellite interferometry to the detection of surface deformation in the Salton Sea geothermal field, California. Geothermal Resources Council Transactions, 33:315-319.

Eneva, M., Falorni, G., Teplow, W., Morgan, J., Rhodes, G., and Adams, D. (2011). Surface deformation at the San Emidio geothermal field, Nevada, from satellite radar interferometry. Geothermal Resources Council Transactions, 35:1647-1653.

Fairley, J. (2009). Modeling fluid flow in a heterogeneous, fault-controlled hydrothermal system. Geofluids, 9(2):153-166.

Fairley, J. P. and Hinds, J. J. (2004). Rapid transport pathways for geothermal fluids in an active Great Basin fault zone. Geology, 32(9):825-828.

Faulds, J. E., Coolbaugh, M. F., Benoit, D., Oppliger, G., Perkins, M., Moeck, I., and Drakos, P. (2010). Structural controls of geothermal activity in the Northern Hot Springs Mountains, Western Nevada: The tale of three geothermal systems (Brady’s, Desert Peak, and Desert Queen). Geothermal Resources Council Transactions, 34:675-683.

Faulds, J. E., Hinz, N. H., Coolbaugh, M. F., Cashman, P. H., Kratt, C., Dering, G., Edwards, J., Mayhew, B., and McLachlan, H. (2011). Assessment of favorable structural settings of geothermal systems in the Great Basin, Western USA. Geothermal Resources Council Transactions, 35:777-783.

Feigl, K. L. and Thurber, C. H. (2009). A method for modelling radar interferograms without phase unwrapping: application to the M 5 Fawnskin, California earthquake of 1992 December 4. Geophysical Journal International, 176(2):491-504.

Fialko, Y. and Simons, M. (2000). Deformation and seismicity in the Coso geothermal area, Inyo County, California: Observations and modeling using satellite radar interferometry. Journal of Geophysical Research, 105(B9):21781-21793.

Foxall, B. and Vasco, D. (2003). Inversion of synthetic aperture radar interferograms for sources of production-related subsidence at the Dixie Valley geothermal field. In Proc. 28th Workshop on Geothermal Reservoir Engineering (Stanford University, CA).

Gesch, D., Oimoen, M., Greenlee, S., Nelson, C., Steuck, M., and Tyler, D. (2002). The national elevation dataset. Photogrammetric engineering and remote sensing, 68(1):5-32.

Goldstein, R. M. and Werner, C. L. (1998). Radar interferogram filtering for geophysical applications. Geophysical Research Letters, 25(21):4035-4038.

Hole, J., Bromley, C., Stevens, N., and Wadge, G. (2007). Subsidence in the geothermal fields of the Taupo Volcanic Zone, New Zealand from 1996 to 2005 measured by InSAR. Journal of volcanology and geothermal research, 166(3):125-146.

Jolie, E., Faulds, J., and Moeck, I. (2012). The development of a 3D structural-geological model as part of the geothermal exploration strategy - a case study from the Brady’s geothermal system, Nevada, USA. In Proc. 37th Workshop on Geothermal Reservoir Engineering (Stanford University, CA).

Jonsson, S. (2009). Anthropogenic and natural deformation on Reykjanes Peninsula, southwest Iceland, observed using InSAR time-series analysis 1992-2008. In EGU General Assembly Conference Abstracts, volume 11, page 5142.

Kirkpatrick, S., Gelatt Jr, C. D., and Vecchi, M. P. (1983). Optimization by simulated annealing. Science, 220(4598):671-680.

Ali et al.

12

Malvern, L. E. (1969). Introduction to the Mechanics of a Continuum Medium. Prentice-Hall, Englewood Cliffs, NJ.

Massonnet, D. and Feigl, K. L. (1998). Radar interferometry and its application to changes in the earth’s surface. Reviews of Geophysics, 36(4):441-500.

Massonnet, D. and Rabaute, T. (1993). Radar interferometry: limits and potential. Geoscience and Remote Sensing, IEEE Transactions on, 31(2):455-464.

Masters, E. (2011). Interferometric synthetic aperture radar analysis and elastic modeling of deformation at the Svartsengi geothermal field in Iceland, 1992-2010: Investigating the feasibility of evaluating reservoir pressure changes from low earth orbit. Master’s thesis, University of Wisconsin, Madison, WI.

Mogi, K. (1958). Relations between the eruptions of various volcanoes and the deformations of the ground surfaces around them. Bull. Earthquake Res. Inst., 36:99-134.

Moulding, A. and Brikowski, T. (2012). Three dimensional modeling of Basin and Range geothermal systems using TOUGH2-EOS1SC. In Proc. TOUGH Symposium (Lawrence Berkeley National Laboratory, Berkeley, CA).

Okada, Y. (1985). Surface deformation due to shear and tensile faults in a half-space. Bulletin of the Seismological Society of America, 75(4):1135-1154.

Oppliger, G., Coolbaugh, M., and Foxall, W. (2004). Imaging structure with fluid fluxes at the Bradys geothermal field with satellite interferometric radar (InSAR): New insights into reservoir extent and structural controls. Geothermal Resources Council Transactions, 28:37-40.

Oppliger, G., Coolbaugh, M., and Shevenell, L. (2006). Improved visualization of satellite radar InSAR observed structural controls at producing geothermal fields using modeled horizontal surface displacements. GRC Trans, 30:927-930.

Oppliger, G., Coolbaugh, M., Shevenell, L., and Taranik, J. (2005). Elucidating deep reservoir geometry and lateral outflow through 3-D elastostatic modeling of satellite radar (InSAR) observed surface deformations. Geothermal Resources Council Transactions, 29:419-424.

Sandwell, D. T. and Price, E. J. (1998). Phase gradient approach to stacking interferograms. Journal of Geophysical Research, 103(B12):30183-30204.

Sarychikhina, O., Glowacka, E., and Mellors, R. (2007). Preliminary results of a surface deformation study, using differential InSAR technique at the Cerro Prieto geothermal field. BC, México: Geothermal Resources Council Transactions, 31:581-584.

Shevenell, L., Oppliger, G., Coolbaugh, M., and Faulds, J. (2012). Bradys (Nevada) InSAR anomaly evaluated with historical well temperature and pressure data. Geothermal Resources Council Transactions, 36:1383-1390.

Smith, I. M. and Griffiths, D. V. (2005). Programming the finite element method. Wiley.

Vasco, D., Rutqvist, J., Ferretti, A., Rucci, A., Bellotti, F., Dobson, P., Oldenburg, C., Garcia, J., Walters, M., and Hartline, C. (2013). Monitoring deformation at the Geysers geothermal field, California using C-band and X-band interferometric synthetic aperture radar. Geophysical Research Letters.

Wang, H. F. (2000). Theory of linear poroelasticity: With applications to geomechanics and hydrogeologoy. Princeton University Press.

White, J. A. and Borja, R. I. (2008). Stabilized low-order finite elements for coupled solid-deformation/fluid-diffusion and their application to fault zone transients. Computer Methods in Applied Mechanics and Engineering, 197(49):4353-4366.

Wicks, C. W., Thatcher, W., Monastero, F. C., and Hasting, M. A. (2001). Steady state deformation of the Coso Range, east central California, inferred from satellite radar interferometry. Journal of Geophysical Research, 106(B7):13769-13780.

Zheng, Y., Burridge, R., Burns, D. R., et al. (2003). Reservoir simulation with the finite element method using biot poroelastic approach. Technical report, Massachusetts Institute of Technology. Earth Resources Laboratory.

Ali et al.

13

Table 1 List of interferometric pairs analyzed in this study. The term Ha denotes the altitude of ambiguity (Massonnet and Rabaute, 1993).

Pair Sat. Track Ha(m) First epoch (ti) Second epoch (tf) Span Orbit Year Orbit Year (days)

1 TDX1 167 -105.20 15167 2013.2027 17004 2013.5342 121 2 TDX1 167 401.02 14499 2013.0822 33400 2013.4740 143 3 TSX1 536 485.79 30948 2013.0301 33286 2013.4521 154 4 TSX1 91 187.07 28147 2012.5273 30652 2012.9781 165 5 TDX1 167 -268.05 14499 2013.0822 17004 2013.5342 165 6 TSX1 536 135.24 29779 2012.8197 32785 2013.3616 198 7 TSX1 536 -154.71 29946 2012.8497 32952 2013.3918 198 8 TDX1 167 -245.00 12996 2012.8115 16169 2013.3836 209 9 TSX1 536 137.35 29946 2012.8497 33286 2013.4521 220

10 TSX1 536 -484.01 29779 2012.8197 33286 2013.4521 231 11 TSX1 536 -519.71 25771 2012.0984 29779 2012.8197 264 12 TDX1 167 143.13 12996 2012.8115 17004 2013.5342 264 13 TSX1 536 2269.95 25103 2011.9781 29779 2012.8197 308 14 TSX1 536 -112.28 25103 2011.9781 29946 2012.8497 319 15 TSX1 536 -165.32 25771 2012.0984 30948 2013.0301 341 16 ERS2 256 110.87 12134 1997.6192 17144 1998.5781 350 17 TSX1 536 -271.44 25103 2011.9781 30948 2013.0301 385 18 ERS2 256 803.14 12134 1997.6192 17645 1998.6740 385 19 TSX1 167 1785.01 24716 2011.9096 30728 2012.9918 396 20 TSX1 167 -236.03 24716 2011.9096 31396 2013.1123 440 21 ERS2 256 -238.69 42695 2003.4658 49208 2004.7104 455 22 TSX1 536 182.81 25771 2012.0984 32785 2013.3616 462 23 TSX1 167 -291.99 25718 2012.0902 33066 2013.4137 484 24 TSX1 536 -250.61 25771 2012.0984 33286 2013.4521 495 25 TSX1 167 -182.12 23881 2011.7589 31396 2013.1123 495 26 TSX1 167 1132.06 24716 2011.9096 32231 2013.2630 495 27 TSX1 536 127.63 25103 2011.9781 32785 2013.3616 506 28 TSX1 536 -615.18 25103 2011.9781 33286 2013.4521 539 29 TSX1 167 -2696.73 23881 2011.7589 32231 2013.2630 550

Ali et al.

14

Table 2 Parameters estimated for the elastic model consisting of two point sources (Mogi, 1958) and two, mode 1, rectangular tensile dislocations (Okada, 1985) in halfspace, using the ensemble dataset.

Parameter Value Uncertainty Mogi1 Easting (m) 328193.80 72.90 Mogi1 Northing (m) 4408507.00 102.51 Mogi1 Depth (m) 327.55 81.45 Mogi1 Volume Increase (m/yr) -7305.14 3851.16 Mogi2 Easting (m) 327341.20 121.93 Mogi2 Northing (m) 4406495.00 110.04 Mogi2 Depth (m) 414.76 138.00 Mogi2 Volume Increase (m/yr) -6540.81 8250.19 Okada1 Length (m) 454.20 183.62 Okada1 Width (m) 13.07 15.42 Okada1 Depth (m) 101.55 60.33 Okada1 Negative Dip (deg) -9.58 11.55 Okada1 Strike CCW from N (deg) 8.72 8.07 Okada1 Easting (m) 328294.70 43.55 Okada1 Northing (m) 4408090.00 48.50 Okada1 Tensile Opening (m) -0.08 0.12 Okada2 Length (m) 1587.45 386.27 Okada2 Width (m) 19.72 12.46 Okada2 Depth (m) 351.04 73.02 Okada2 Negative Dip (deg) -8.92 9.19 Okada2 Strike CCW from N (deg) 16.06 6.07 Okada2 Easting (m) 327491.50 43.00 Okada2 Northing (m) 4406715.00 72.18 Okada2 Tensile Opening (m) -0.57 0.25 Rate of Total Volume Increase (m/yr) -27746.43 811.82