Hendra Grandis 1 and Prihadi Sumintadireja 2

A BRIEF REVIEW FOR THE PROPER APPLICATION OF MAGNETOTELLURIC (MT) AND CONTROLLED-SOURCE AUDIO-FREQUENCY MAGNETOTELLURIC (CSAMT) IN GEOTHERMAL EXPLORATION Hendra Grandis 1 and Prihadi Sumintadireja 2 ABSTRACT (1) Applied Geophysics Research Division Faculty of Mining and Petroleum Engineering Institut Teknologi Bandung e-mail: grandis@geoph.itb.ac.id, grandis@earthling.net (2) Applied Geology Research Division Faculty of Earth Science and Technology Institut Teknologi Bandung e-mail: prihadi@gc.itb.ac.id Magnetotelluric (MT) method is a frequency domain sounding technique that uses naturally existing electromagnetic (EM) fields as the source to infer the subsurface resistivity distribution. Controlled- Source Audio-frequency Magnetotellurics (CSAMT) is derived from MT and uses an artificial EM source (typically grounded electric dipole) in the range 0.1 Hz to 10 khz. Both MT and CSAMT are well recognized as geophysical methods in geothermal exploration since both results in resistivity structure that is more or less related to the temperature regime of the subsurface. However, from empirical point of view sometimes one method (e.g. CSAMT) delivers superior results than the other method (e.g. MT) in geothermal exploration, and vice versa. Successful applications of one technique were then used to penalize the other technique, especially in Indonesia. This situation is mainly due to lack of knowledge of the basic concept of both MT and CSAMT and also their processing and modeling aspects. Unfortunately, this point of view influences policy and decision making in adopting one method by disregarding the other in the exploration program. This paper describes both MT and CSAMT techniques from theoretical and practical perspectives in order to underline the possible difference of outcomes when they are applied in geothermal exploration. Synthetic data as the response of simple models as well as real (field) data are used to exemplify pitfalls that may generate misleading interpretations. We hope that in the future any exploration program employing geophysical methods, especially MT or CSAMT will be based on correct and well-founded scientific reasoning related to specific target of the survey. INTRODUCTION In geothermal exploration, electrical geophysical methods are of prime importance due to the close and direct relationships between temperature and resistivity of materials. Hydrothermal fluids and also their overlying cap rock have large resistivity contrasts relative to their host environment. Magnetotellurics (MT) and its controlled-source, high-frequency complement CSAMT are the main tools to delineate such geological features and the basic structural controls based on their resistivity expression. Case histories have shown that both MT (e.g. Newman et al., 2008; Cumming and Mackie, 2010) and CSAMT (e.g. Wannamaker, 1997) have been successful in establishing geothermal resources. There were less-successful results from MT and CSAMT as well but with very limited or even no publication at all. Successful applications of one technique were then used to penalize the other technique. Unfortunately, this point of view influences policy and decision making in adopting one method by disregarding the other in the geothermal exploration program. This situation is mainly due to lack of knowledge of the basic concept of both MT and CSAMT and also their processing and modeling aspects. This paper describes both MT and CSAMT techniques from theoretical and practical perspectives in order to underline the possible difference of outcomes when they are applied in

geothermal exploration. Synthetic data as the response of simple models as well as real (field) data are used to exemplify pitfalls that may generate misleading interpretations. MT AND CSAMT METHODS Both MT and CSAMT are frequency domain sounding techniques employing variations of electromagnetic (EM) fields as the primary source. Orthogonal components of electric (E x, E y ) and magnetic (H x, H y ) fields are recorded as time series data and then processed by using spectral analysis that results in complex frequencydependent impedance (Z). Impedance is the ratio of the electric and magnetic fields, from which the information on the subsurface can be inferred. For a complete set of orthogonal components, electric and magnetic fields are related by the impedance tensor (Z), E E x y Z = Z xx yx Z Z xy yy H H x y or E= ZH (1) MT and CSAMT sounding data, i.e. apparent resistivity (ρ a ) and phase (φ) as function of period (T) or frequency (f = T -1 ) can be calculated from each component of the impedance tensor (Z ij ), 1 ρ ; 2 ( a) ij = Zij ωµ 0 φ ij Im Z ij = tan 1 (2) Re Zij where ω = 2πf, µ 0 = 4π 10-7 (in SI unit) is freespace magnetic permeability. Data along a profile line can also be presented as pseudo-section, i.e. apparent resistivity and phase contour plot with distance on the horizontal scale and period or frequency on the vertical scale. The increasing period (or decreasing frequency) vertically downward represents increasing depth. It allows apparent resistivity and phase at many MT or CSAMT stations to be shown in a single figure and facilitates preliminary and qualitative interpretation. Quantitative interpretation can only be performed after modeling of the data using 1-D, 2-D or 3-D models to represent the subsurface resistivity distribution. Multi-dimensional MT and CSAMT modeling necessitates complete impedance tensor, while scalar impedance may result in 1-D model where resistivity varies only with depth. The investigation depth for both MT and CSAMT is usually expressed as the skin depth (in meter), δ 500 ρt or ρ δ 500 (3) f where ρ is resistivity of the homogeneous medium equivalent (see Figure 1). The skin depth serves as an approximation only and it tends to overestimate the investigation depth of an EM method. The lower frequency for CSAMT is in the range of 0.1-10 Hz (or T between 0.1-10 sec.). With longer periods reaching 100 sec. or more, MT investigation depth is significantly deeper than CSAMT. Typically, CSAMT is for the target depth around 1-2 km while MT is for 4-5 km. SPECIFIC CHARACTERISTICS OF CSAMT Near-Field Effect By using a transmitter (usually a grounded electric dipole, see Figure 2) to generate the source fields, CSAMT method overcomes problems associated with weak natural EM fields used in conventional audio-frequency MT surveys. However, in addition to survey logistic difficulties in providing high-powered transmitter, there are effects related to the proximity of the signal source. The assumption of uniform plane-waves as in MT implying that the wave s source is distant, is no longer valid in CSAMT. With finite transmitter-receiver distance (R), CSAMT data are likely to contain far-field and near-field data, separated by a transition zone. Generally, the near-field is defined as the zone where the separation between the transmitter and receiver is less than 3 times the skin depth for a uniform, isotropic material underlying the zone (Zonge and Hughes, 1991). The far-field is where the MT approximation is valid. Figures 3 and 4 show comparisons of CSAMT and MT sounding curves for two three-layered 1-D synthetic models with R = 2 km. In the farfield zone apparent resistivities for CSAMT and MT are identical up to a certain period. Beyond that zone, near-field zone prevails. The near-field zone for Model-1 starts at T = 0.1 sec., while in the Model-2 it starts at shorter period, i.e. approximately 0.002 sec. These facts confirm that for a single transmitter-receiver separation in CSAMT sounding site, division of the data into near-field

and far-field depends on the skin depth. This in turn is function of the period and resistivity of the medium. The combination of the first two layers in Model-1 (1000 Ohm.m and 10 Ohm.m) are significantly lower than those of the Model-2 (100 Ohm.m and 1000 Ohm). Therefore, the far-field (R >> δ) extends to longer period in Model-1 than in the Model-2. From Figures 3 and 4 it is also obvious that in the near-field zone, the CSAMT apparent resistivity curves do not make an asymtote to the resistivity of lowest layer as in MT. Therefore, qualitative interpretation of apparent resistivity pseudosection and even quantitative 1-D modeling using MT approximation are definitely misleading. Yamashita et al. (1985) and Bartels and Jacobson (1987) advocate the correction of the near-field data such that CSAMT data can be interpreted using MT modeling tools. Grandis (2000) presented in detail the implementation of such correction. Apparent resistivity for the near-field is calculated by using a formula taking into account the effects of array geometry (Zonge and Hughes, 1991), i.e. R ( ρ a) ij = Zij or ( ρ a) ij = kn R Zij (4) 2 where R is the trasmitter-receiver distance and k n is the generalized multiplication factor for the impedance to obtain the apparent resistivity. In the transition zone from the near-field side, k n is interpolated from coefficients determined from a generalization of the homogeneous medium case. Similar approach is employed in the transition zone from the far-field side. Grandis (2000) proposed the normalized frequency for easier identification of the transition zone. Figure 5 shows the correction of the synthetic data presented in Figures 3 and 4. Corrected CSAMT data are fairly similar with the synthetic MT data, especially when the near-field effect is not too severe as in Model-1 (see Figure 5). Application of the correction technique for the field CSAMT data from Kamojang geothermal field is presented in Figure 6. Qualitative interpretation of the uncorrected pseudo-section would under-estimate the low resistivity layer. With relatively dense sounding points (100 m apart), the lateral boundary at the profile's right-end would also be missed. Source Effects Zonge and Hughes (1991) and Mitsuhata (2000) discuss the effects of the artificial source in CSAMT by using 2-D numerical model studies. The source overprint effect is illustrated in Figure 7 where geological environment near the transmitter affects CSAMT data. In addition to the subsurface resistivity distribution under the sounding point, CSAMT data also reflect the one under the transmitter. The shadow effect is another source related problems in CSAMT. Figure 8 shows the apparent resistivity pseudo-section of a line perpendicular to the transmitter as a result of 2-D numerical modeling by Zonge and Hughes (1991). Soundings at the line parallel to and on the broadside of the transmitter (for example at 7 km from the transmitter, at the right-most of Figure 8) would reflect the effect of the conductive anomaly located between trasmitter and receiver. Scalar Mode Measurement Complete tensor CSAMT measurement is possible by employing two perpendicular transmitters to provide two independent source polarisations. However, survey logistic difficulties prohibit tensor CSAMT mode for routine operations, except for research and studies for very specific purposes. In most CSAMT surveys, scalar mode measurement is always implemented. Typically, a transmitter is positioned along the x- axis and the survey grid is arranged at the broadside of the transmitter, along a line parallel to the x-axis. A multi-channel receiver records electric field simultaneously at seven dipoles (E x ) and one perpendicular magnetic field (H y ). Due to restriction in cable transmission, the distance between CSAMT soundings is limited to around 100 m (see Figure 2). In the processing stage, each electric field from a single set-up is paired with associated magnetic field from the same set-up. The results are scalar impedances Z xy = E x /H y from which apparent resistivity and phase can be calculated. Owing to redundant and spatially high density data, apparent resistivity pseudo-section may appear as a high-resolution image of the subsurface. Even so, modeling of scalar impedance data from each sounding point can only be performed by using 1-D approximation of the subsurface.

DISCUSSION CSAMT technique was devised to overcome disadvantages of the natural field mode, i.e. MT method. Both MT and CSAMT methods result in resistivity estimation of the subsurface. There is no reason to prefer one method over the other based only on empirical evidences related to success or failure of one of them in determining geothermal resources. CSAMT method is the most appropriate if all the following conditions are true: (i) target is relatively shallow, (ii) the study is intended as a preliminary or reconnaissance survey, (iii) the survey area suffers from both cultural and natural noise, and (iv) accessibility of survey area does not prohibit the use of high-powered electrical generator for the transmitter. For deeper target and for advanced stage of the geothermal exploration where detailed subsurface resistivity distribution shall be achieved, then MT method is more suitable. MT data quality can be improved by employing Remote Referenced MT (RRMT) measurement (e.g. Krings et al., 2007). Combination of MT and AMT (Audio-frequency MT) can still be considered to achieve an efficient compromise between delineation and detailed purposes. This is without sacrificing the completeness of the data by measuring complete tensor impedance even for AMT frequency band. In such case, dimensionality analysis can be performed on the tensor impedance data. When CSAMT is chosen, the transmitter should be placed as far as possible from the sounding points to obtain only far-field data. Far from the transmitter, the electromagnetic fields are planar, and CSAMT data can be interpreted using conventional approaches developed for MT data. Contrary to the previous requirement, there is a need to minimize transmitter-receiver separation for maximum signal. Therefore, most CSAMT data sets will contain some near-field data. However, if most of the data are far-field, then we can discard a small number of data in the nearfield and consider far-field data only in all subsequent CSAMT data analysis. Cautious CSAMT data analysis and interpretation must be performed, when they contain mostly near-field data. Although the validity of near-field CSAMT data correction is still debatable, at least it can produce better and more "correct" pseudosections. Rather than interpreting directly apparent resistivity pseudo-sections from uncorrected CSAMT data, the use of existing MT 1-D modeling to near-field corrected data is advisable. Nowadays, CSAMT 1-D inversion is available employing full-solution of the EM fields due to artificial electrical line source (e.g. Routh and Oldenburg, 1999; Wang et al., 2012). In such case, there is no need for near-field correction of CSAMT data. There are also methods for CSAMT 2-D inversion (e.g. Mitsuhata, 2000; Wang et al., 2009). However, such 2-D inversion necessitates CSAMT tensor or at least vector impedance data, both resulting from measurement using two perpendicular sources. For more practical purposes, we can consider that the profile line is oriented across a presumed geologic strike, then the scalar impedance Z xy = E x /H y may be regarded as TM-mode data. Such "incomplete" data can be inverted using available codes or softwares for MT 2-D inversion (e.g. Uchida, 1993). Figure 9 presents the result of 2-D inversion of scalar CSAMT (far-field data only) from Lahendong geothermal field. The geological interpretation of the model is beyond the scope of this paper. However, it can be observed that the 2-D subsurface resistivity distribution is very informative. Combined with results from other profiles, the model may generate meaningful interpretation related to the local geology of the survey area. CONCLUSION There is no doubt that both MT and CSAMT play significant role as main geophysical methods in geothermal exploration. Both techniques are capable to produce resistivity structure that is more or less related to the temperature regime of the subsurface. As in any other geophysical methods, data analysis and interpretation shall be performed with great cautions. In MT, the low signal to noise ratio demands data acquisition and processing that can minimize the influence of noise. The term Robust Remote Referenced MT (RRRMT) has become a standard practice in MT. The complete tensor impedance data shall be exploited as far as possible to extract

the dimensionality of the structures. It is wellknown that most geothermal fields related to volcanic area have very complex structures. In this regard, the use of the tensor decomposition technique is therefore essential (see Weckmann et al., 2003 for a relatively recent review). Multidimensional MT modelings (2-D or even 3-D) are also of interest in order to take into account the complicated subsurface structure of a geothermal prospect. The advantage of CSAMT, in addition to high signal to noise ratio, is the redundancy of the data along a profile line. The CSAMT appears to provide laterally high-resolution image of the subsurface. However, CSAMT data and therefore its pseduo-section suffer from erroneous effects due to finite transmitter-receiver separation. Therefore, direct qualitative interpretation from uncorrected CSAMT data is not advisable. Existing full-solution for CSAMT modelings in 1-D and 2-D are advantageous for CSAMT interpretation. These exonerate the need for nearfield correction. However, the fact that most CSAMT data are scalar prohibit us to fully take advantage from such modelling tools. For most practical use of CSAMT, it would be sufficient to consider the existing tools at hands. First, perform CSAMT only if all requirements described in the discussion above are fulfilled. Second, select only far field data or apply nearfield correction, then qualitatively analyse the resulting pseudo-sections. Third, perform MT 1-D and/or 2-D (with TM-mode data only) inversion modeling to the far-field data or near-field corrected CSAMT data. REFERENCES Bartel, L. C., and R.D. Jacobson, 1987, Results of a controlled-source audio frequency magnetotelluric survey at the Puhimau thermal area, Kilauea Volcano, Hawaii, Geophysics, 52, 665-677. Cumming, W., and R. Mackie, 2010, Resistivity imaging of geothermal resources using 1D, 2D and 3D MT inversion and TDEM static shift correction illustrated by a Glass Mountain case history, Proceedings World Geothermal Congress 2010 Bali, Indonesia. Grandis, H., 2000, Koreksi efek sumber pada data Controlled-Source Audio-Magnetotellurics (CSAMT), Jurnal Teknologi Mineral, VII/1, 43-50. Krings, T., O. Ritter, G. Munoz and U. Weckmann, 2007, MT robust remote reference processing revisited, 22 Kolloquium Elektromagnetische Tiefenforschung, Czech Republic. Mitsuhata, Y., 2000, 2-D electromagnetic modeling by finite-element method with a dipole source and topography, Geophysics, 65, 465-475. Newman, G.A., E. Gasperikova, G.M. Hoversten and P.E. Wannamaker, 2008, Three-dimensional magnetotelluric characterization of the Coso geothermal field, Geothermics, 37-4, 369-399. Routh, S.P., and D.W. Oldenburg, 1999, Inversion of controlled-source audio frequency magnetotelluric data for a horizontally layered earth, Geophysics, 64, 1689-1697. Uchida, T., 1993, Smooth 2-D inversion of magnetotelluric data based on statistical criterion ABIC, Journal of Geomagnetism & Geoelectricity, 45, 841-858. Wang, R., M. Wang, Q.Y. Di and G.J. Wang, 2009, 2- D numerical study on the effect of conductor between the transmitter and survey area in CSEM exploration, Applied geophysics, 6, 311-318. Wang, R., C. Yin, M. Wang, and G. Wang, 2012, Simulated annealing for controlled-source audiofrequency magnetotelluric data inversion, Geophysics, 77, E127-E133. Wannamaker, P., 1997, Tensor CSAMT survey over the Sulphur Springs thermal area, Valles Calder, New Mexico, USA, Part 1: Implications for structure of the western Caldera, Geophysics, 62, 451-465. Weckmann, U., O. Ritter and V. Haak, 2003, Images of the magnetotelluric apparent resistivity tensor, Geophysical Journal International, 155, 456-468. Yamashita, M., P.G. Hallof and W.H. Pelton, 1985, CSAMT case histories with a multi-channel CSAMT system and discussion of near-field data correction, 55th SEG Annual Meeting Expanded Abstracts. Zonge, K.L., and L.J. Hughes, 1991, Controlled-source audio-frequency magnetotellurics, in Nabighian, M.N., Ed., Electromagnetic Methods in Applied Geophysics, Vol. 2, Society of Exploration Geophysicists.

Figure 1. Skin depth for MT and CSAMT methods as function of the period for different homogeneous medium equivalent. Figure 3. MT and CSAMT apparent resistivity sounding curves (top) and phase (bottom) for the synthetic Model-1. Figure 2. Typical lay-out of transmitter (Tx) and receiver (R) in a CSAMT survey. The survey grid is located at the broad-side of the transmitter. The scalar mode measurement at each sounding point is composed by a single component of electric and magnetic fields. Figure 4. MT and CSAMT apparent resistivity sounding curves (top) and phase (bottom) for the synthetic Model-2.

Figure 5. Comparison of corrected CSAMT and MT apparent resistivity sounding curves for Model-1 (left) and Model-2 (right). Figure 6. Apparent resistivity pseudo-section of CSAMT data from one line at Kamojang geothermal prospect, uncorrected (top) and corrected (bottom) using the method described by Grandis (2000). Figure 7. Apparent resistivity sounding curves for different geological environment at the transmitter location illustrating the source overprint effect modeled by Zonge and Hughes (1991).

PROCEEDINGS The 12TH ANNUAL INDONESIAN GEOTHERMAL ASSOCIATION MEETING & CONFERENCE pseudo section of a line perpendicular to the transmitter as a result of 2-D Figure 8. Apparent resistivity pseudo-section numerical modeling by Zonge and Hughes (1991). Soundings at the line parallel to and on the broad broad-side of the transmitter would reflect the effect of the conductive conductive anomaly located between trasmitter and receiver. Figure 9. Result of 2-D MT (TM-mode) (TM inversion modeling of scalar CSAMT far-field far data only from Lahendong geothermal field.