U.S.

DEPARTMENT OF ENERGY GEOTHERMAL TECHNOLOGIES PROGRAM ENHANCED GEOTHERMAL SYSTEMS PEER REVIEW REPORT Project Title: Imaging Multi-Dimensional Electrical Resistivity Structure as a Tool in Developing Enhanced Geothermal Systems CPS Identifier: 17468-11961 Principal Investigator: Philip E. Wannamaker Sponsoring Organization: University of Utah, Energy & Geoscience Institute 423 Wakara Way, Suite 300 Salt Lake City, UT 84108 Other Investigators: Yutaka Sasaki (Kyushu University), Fang Sheng (Baker Atlas Inc.), Virginia Maris (U Utah Ph.D. graduate student)

Project Purpose Project aims to ultimately increase economically viable geothermal resources in the U.S. and reduce the levelized price of geothermally-derived electricity by advancing the ability to image and interpret underground geophysical properties related to fluid flow in enhanced (engineered) geothermal systems (EGS). Since an increased fluid content due to fracturing, and the development of more conduc tive alteration minerals (clays, etc.) as a result of fluid flow, can give rise to an electrical resistivity contrast, electromagnetic (EM) methods of probing have been investigated and applied for many years. Development in this proposal of three-dimensional inversion capability for complementary EM/galvanic (DC) resistivity data sets which can run on serial PCs or low-cost clusters is intended to provide a tool for fracture zone mapping, and the evaluation of subsurface processes affecting fluid flow. It is also intended to provide technology pertinent to private industry, the contract surveying arm of which has taken the responsibility for EM/DC data collection whether for EGS purposes or more general exploration goals. Project Objective(s) The 3-D inversion development herein addresses the principal facets of practical subsurface imaging technology: maximal solution stability, computational accuracy, high speed, minimal storage, and hardware affordability. A principal goal is to bring meaningful 3-D inversion capability to the PC environment. Given advances in computer platforms, and the development proposed here, the Gauss-Newton (G-N) parameter step known for its good convergence now appears practical for the largest data sets encountered so far in geothermal applications. Computational accuracy is ensured through use of a particular staggered grid implementation shown to have superior characteristics to others published so far. The chief bottleneck of the G-N approach has been laborious computation of the parameter sensitivities in perhaps the most novel development of the project effort, this is overcome through application of the integral equations method. Storage, and run-time, are reduced further by porting the algorithm to an inexpensive but powerful Beowulf-type computer cluster, a hardware platform which is affordable by small companies. EGS Peer Review Paper, Wannamaker Page 1

Four data sets will be analysed using this inversion capability. The first will be the extensive collection of MT measurements (>100 stations) conducted at the Coso, CA, EGS site in 2003-2005. These results were taken mainly over the east flank region of high temperatures but limited permeability, which is the subject of an ongoing EGS project (P.I. P. Rose, DE-PS07-00ID13913). The second data set is a group of about 65 evenly-distributed MT soundings over the Karaha-Telaga Bodas andesitic system. The third set will be a multi-profile, CSAMT set from the Surprise Valley/Lake City geothermal prospect in northeastern California. This test will examine the ability of explicit accommodation of finite source effects to significantly extend the depth of exploration (hopefully by about a factor of three). Finally, a full multi-profile, simultaneous MT/DC array data set from the Battle Mountain area of northwestern Nevada will be inverted as a test of complete imaging capability using such a complementary electrical data set. From the standpoint of the EGS draft multi-year plan, it is important to characterize existing porosity, permeability and alteration in an existing geothermal field in order to more accurately site injection/stimulation wells to expand field production. Existing Coso results presented shortly exemplify this. In addition, the approach taken here allows inversion for changes from an initial state, such as may be pertinent to EM monitoring surveys performed during and following large-scale stimulation experiments or over the fields production history. So that MT resistivity analysis is not a bottleneck in the EGS evaluation process and to substantially expand the number of parties capable of carrying out 3-D analysis efficiently, development of inversion techniques workable on modern PCs or small clusters is an important goal. It is our intention to have a public domain, efficient and stable 3-D inversion platform by the close of the project. Duration: October 1, 2004 to September 30, 2007 Funding Fiscal Year 2005 2006 2007 DOE ($k) 208 149 142 Funding Cost Share ($k) 58 70 124 Total 266 219 266

Plans and Approach The 3-D inversion development proposed herein for efficient imaging of geothermal systems addresses the principal facets of practical subsurface imaging technology: maximal solution stability, computational accuracy, high speed, minimal storage, and hardware affordability. Both plane-wave and finite sources will be considered jointly to accommodate the new-generation systems which have emerged (Figure 1). These systems exploit the complementary resolving power of EM and DC methods to improve both electrical resistivity and induced polarization sections (Madden, 1971, AGU Mono 14). However, separate inversion capability will be created also for the numerous important MT and DC data sets in existence or to be collected.

EGS Peer Review Paper, Wannamaker Page 2

Figure 1. Generalized EM acquisition mode as exemplified by dense MT/DC-IP array system with which we have worked. Contiguous tensor MT data are followed by poledipole DC with roving interior current pole, building up several profiles in 3 -D if desired.

Figure 2. Governing equations for the three principal steps of an inversion process described in text. First portion is finite difference (FD) method, third uses integral equations (IE). The three principal steps in an inversion algorithm are summarized in Figure 2. First is an efficient and accurate forward problem which can handle discretized earth models large and complex enough for practical field data sets (finite differences in our case). Second is an iterative parameter step based usually on maximum likelihood statistics combining misfit between data-model responses and a model stabilizer, the latter of which is typically smoothing or adherence to a priori information (e.g., Taranto la, 1987, Elsevier). Whether this estimator is solved directly or through relaxation, it requires the parameter Jacobians, which express the incremental change EGS Peer Review Paper, Wannamaker Page 3

in measured response due to incremental change of a small subsection (parameter) of the earth. This is the third step in the inversion procedure, the efficient solution to which using integral equations , to our knowledge, has not been attempted before. Results This is a new project recently getting underway, and P.I. Wannamaker as well has necessarily been partially involved in other projects including FY04 carryover. Chartfield numbers at the University of Utah for which to charge on this project also were not established until October 31/04. Effort so far in the project has largely been devoted to support the ongoing EGS project at the Coso geothermal field, where MT measurements in support of reservoir understanding and hydraulic stimulation have been carried out by the P.I. (Wannamaker et al., 2004, Proc. SGP). In addition to ~100 separated five-channel sites under DOE and US Navy support, a dense MT array line of 52 contiguous bipoles was measured across the northern east flank ~400 m north of experimental well 34-9RD2. An inversion section for this array line appears in Figure 3, using the a -priori inversion code of the P.I. developed under previous DOE funding.

Figure 3. Two -dimensional inversion section of TM mode of array MT data across the northern east flank of the Coso prospect. Wellheads of 34-A9 and 34-9RD2 project northward to approximately site 021. Low resistivities are plotted with warm colors. Most interesting is a steeply west-dipping conductor in resistive lithology of the east flank, which we interpret to reflect higher fluid, alteration and permeability associated with east flank production. Most of the productive wells of the east flank are west-dipping and may be intersecting this zone. Note that 34-9RD2, up to this point not a producer, may barely graze the zone. In late February/05, deepening of this well commenced in preparation for a massive cold water injection to stimulate permeability. However, several large fractures with attendant lost circulation were encountered upon redrilling, prompting reconsideration of the need for stimulation at present. We view this as confirmation of the MT structure imaged here. Under separate support from the U.S. Navy, Newman et al. (2005, Proc. SGP) have constructed 2-D images from coarse (810 site) subprofiles of the five-channel site collection. Although of much lower

EGS Peer Review Paper, Wannamaker Page 4

resolution, these confirm that the structure persists to the southern end of the current east flank field. These will be shown in the oral presentation. Regarding the inversion development components of the project, in the short time available so far the P.I. has continued to organize program units of the FD MT forward program and the newer grounded bipole source code provided by colleague Y. Sasaki of Kyushu University. Storage arrangement in the IE program has been reworked (laboriously) so that common blocks are removed and replaced by argument passing, to facilitate efficient storage management on the computer cluster. Unnecessary symmetry options have been removed also for simplicity. A shell has been written which is analogous to the authors 2-D inversion program for emplacing the forward, jacobian and parameter step components. A comparison is shown between the P.I.s 3-D IE forward problem and Sasakis FD code for a grounded bipole source and CSAMT responses, such as we will consider in the Surprise Valley area (Figure 4). Good agreement is obtained between the two methods. The FD results are judged to be somewhat more accurate in the case of the lower frequency phase, with the IE behavior perhaps reflecting how the non-uniform strength of the source is manifest in the matrix source vector.

Figure 4. Example CSAMT and MT responses over resistive brick in half-space at 200 Hz (left) and 2 Hz (right). Apparent resistivity and impedance phase are from average impedance determinant. Half-space MT responses of 15 ohm-m in apparent resistivity and 45 degrees in phase are denoted with horizontal lines. Shortly after the start of the project, Wannamaker received a copy from Sasaki of the Fortran source code for his full Gauss-Newton 3-D inversion program (Sasaki, 2004, Earth Planets Space). This algorithm is currently too slow to be practical for industrial use including extensive model testing, but it will be quite valuable for testing and comparison to the platform development of this project. We are extremely fortunate to have this working relationship; obtaining advanced modeling source codes from other workers in the field is a difficult issue generally. An example application of Sasakis algorithm over a modest geothermal MT data set appears in Uchida et al. (2003, 3DEM-III Symposium Adelaide). Impact of Work/Merit An irreplaceable measure of the physical geometry and characteristics of a geothermal reservoir and its fracture network is provided by geophysical imaging of the subsurface properties, of which electrical resistivity has been considered one of the most EGS Peer Review Paper, Wannamaker Page 5

important. A major economic system compone nt of the EGS multiyear plan is the procedure of characterizing the resource, particularly its fracture system, for both drilling and circulation purposes. The numerical development of this proposal may yield a large increase in the usability of 3-D interpretation methods in geothermal and elsewhere. We have seen in the Coso project so far that there is a good correlation between low resistivity structure beneath the East Flank, and production and permeability. The discovery of large fractures off the end of 34-9RD2 seems further confirmation of this. Although the correspondence between resistivity and geothermal potential is often complicated, in systems like Coso which are dominated by plutonic lithologies a resistivity structural image may guide well placement for stimulation and field expansion purposes. Using auxiliary data to estimate high temperatures and favorable local stress conditions for stimulating permeability, electrical resistivity can establish domains of existing permeability to tap into. This project provides state-of-the-art training into MT methods and their application to geothermal systems for Ph.D. student Virginia Maris. This will foster newly skilled personnel for the geothermal field and exploration geophysics generally. Maris received her M.Sc. from the University of Manitoba in 2000 under the direction of Prof. Ian Ferguson. She has experience with a wide range of techniques for environmental and exploration purposes including MT/CSAMT, frequency-domain conductivity meters, DC resistivity, GPR, gravity, magnetics, borehole logging, and shallow seismics. Plans for Completion Work is currently underway to merge aspects of the existing MT and finite source versions of Sasakis forward problems with respect to primary field calculations and handling of topography. Wannamaker will continue implementing the more accurate and versatile finite source components from his long-standing integral equations program, starting with the MT module. These are perhaps unexciting, but necessary steps in this technology development. Deliverables for FY05 are a generalized FD forward problem including arbitrarily layered earth host, and demonstration of the value of inexpensive PC clusters in resistivity inversion using our 2-D platform. Good accuracy has been demonstrated and will be shown for the IE jacobians, but further improvements will be investigated in FY06 by defining anisotropic cell conductivities following methods of Fang Sheng (Wang and Fang, 2001, Geophysics). Fang has provided the fully 3-D generalization of this to Wannamaker. An automated routine to generate the expanding IE cells must be coded, building on our 2-D work. Static effects in the jacobians will be investigated and treated probably by defining the data as log apparent resistivity and impedance phase. The shell incorporating the FD forward problem, IE jacobians and G-N step will be completed. Inversion of the Coso and Karaha MT data will be attempted. In FY07, finite grounded source generalization will be added and inversion will be tested on the Surprise Valley CSAMT and Battle Mountain prospect MT/DC data. Emphasis will be place on writeup of results, documentation of computer code, and completion of the Ph.D. research of Maris.

EGS Peer Review Paper, Wannamaker Page 6