Notice: Wiley Online Library will be unavailable on Saturday 30th July 2016 from 08:00-11:00 BST / 03:00-06:00 EST / 15:00-18:00 SGT for essential maintenance. Apologies for the inconvenience.
 We constructed a three-dimensional conceptual model of a geothermal system on the Caribbean island of Montserrat. The model was generated using magnetotelluric resistivity data, earthquake hypocenter data, and a three-dimensional P wave velocity model, all plotted using a shared geographical reference. The results of the study suggest a high-temperature fracture-controlled geothermal system at the intersection of two faults in the SW of the island. We also present a “prospectivity index” map that represents a proxy of the spatial variation in harvestable heat flux at 1500 m depth. The index is the product of relative permeability around modeled faults and a proxy for the subsurface temperature calculated using P wave velocity anomalies.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The Caribbean island of Montserrat is well known for the Soufrière Hills Volcano (SHV) that has been erupting since 1995 [Kokelaar, 2002]. Similar to other volcanic islands, Montserrat exhibits evidence of geothermal activity [Chiodini et al., 1996] with four buried soufrières in the southeast region of the island near SHV (Figure 1). In 1967–1968, Keith Consulting  drilled 62 shallow groundwater wells across the island. Some results from this work have been re-reported in Principe . The three hottest wells, with temperatures between 46.6°C and 71.1°C, were located near an alkali chloride hot spring called “Hot Water Pond” (HWP). Interest in utilizing renewable geothermal energy to generate electricity has prompted a number of studies [Geotermica Italiana, 1991; Principe, 2008; Younger, 2006; Environmental and Geothermal Services (EGS), 2010].
 Literature reveals two competing conceptual models for the geothermal system that has a surface expression at HWP. One model consists of an upflow of geothermal fluid in the southwest associated with the SHV magmatic system where fluids flow to the west, producing HWP spring. This is the simplest model consistent with data prior to this study. This model conforms to a classic model of a high-temperature system, consisting of acid sulfate geothermal features (the soufrières) above a high-temperature upflow and a more distal alkali chloride outflow (HWP). The other model proposed by Ryan et al.  consists of an upflow farther to the west beneath St. George's Hill (SGH) feeding the outflow tapped by HWP. This model was proposed based on subsurface resistivity patterns in the area consistent with a high-temperature upflow, namely, a domed electrical resistivity low at about 1500 m depth that is consistent with a hydrothermally altered smectite clay cap detected via magnetotelluric (MT) and time domain electromagnetic (TDEM) surveys.
 In this study, we analyze the spatial relationships between the available geophysical data sets including MT [Ryan et al., 2009; Poux and Brophy, 2012], seismic tomography [Shalev et al., 2010], and earthquake (EQ) hypocenters [Aspinall et al., 1998] to determine which conceptual model is best supported by the data. Furthermore, the geophysical data are used to constrain the location of drilling targets in the geothermal system to provide both geothermal power and obtain further information about the structure of the geothermal system.
2 Data and Analysis
 As part of an effort to determine feasibility of geothermal exploitation, MT and TDEM resistivity surveys were conducted in 2009 [Ryan et al., 2009]. The main result from this survey was a cross section running approximately east–west through the geothermal system, identifying what was interpreted as a smectite clay cap (2–10 Ωm), with an intermediate resistivity (10–70 Ωm) near vertically aligned zone beneath. This zone was located between SGH and Garibaldi Hill (GH) and surrounded by higher-resistivity (300–2000 Ωm) material. The higher-resistivity zone was interpreted as minimally altered igneous material. The intermediate-resistivity zone was interpreted as fluid-bearing, fractured rock containing high-temperature hydrothermal alteration clay minerals, such as illite, chlorite, and epidote. The cross section is shown in Figure 2.
 A P wave seismic tomography of the island was interpreted by Shalev et al. . Structures to a maximum depth of 5 km were resolved. High P wave velocity (Vp) anomalies were detected beneath the three main volcanic centers on the island, and three slow Vp anomalies were detected on the flanks of the volcanic centers. The anomaly in the SW is roughly congruent with the suspected location of the geothermal system (see supplementary materials for map of Vp anomalies).
 EQ swarms with hypocenters beneath SGH occurred between 1995 and 1997 [Aspinall et al., 1998] in the vicinity of the geothermal system. A study of all located EQs detected by the Montserrat Volcano Observatory network showed that virtually all local seismicity has been located beneath the SHV since 1997.
 We expect to see variation in Vp across the geothermal system since Vp varies with temperature, degree of alteration, and porosity [Kristinsdóttir, et al., 2010; Han et al., 1986]. We also expect to see EQ hypocenters related to fractured fault zones, which create permeable, fluid-conducting structures in the geothermal system.
3 Three-Dimensional Model
 Utilizing multiple geophysical techniques and constraining interpretations with other geological information help build more robust conceptual models. Determining the detailed spatial relationships between multiple independently acquired (3D) three-dimensional geophysical data sets is challenging. To address this challenge, a three-dimensional model comprising all aforementioned data sets (E–W oriented MT cross-section, Vp anomaly data, and earthquake locations in the region below SGH) was constructed using the open-source ParaView (http://paraview.org) data analysis and visualization application. The model was used to observe patterns and correlations between the data sets to constrain geological interpretation. A digital elevation model (DEM) (NASA, Jet Propulsion Laboratory, http://asterweb.jpl.nasa.gov/gdem.asp) was included to aid in delineating surface features. To estimate fault geometry beneath the SGH, we found a best fit plane to the hypocenter cloud using principal component analysis (PCA). The plane had a strike of 57° and a near vertical dip of approximately 89.5°. HWP is associated with Sand Ghaut fault which has a measured strike of 50° [EGS, 2010]. The strike of the best fit fault plane is less well constrained than its dip because of the limited lateral spread of hypocenters. We have therefore altered the strike of the plane in the model to 50° on the assumption that it is associated with Sand Ghaut fault. The Vp anomaly data are a visualization of the percentage deviation from the starting Vp model. In Figure 2, the −6% isosurface is shown. All data were referenced to the DEM. An animation of the rotating 3D model is included in the supplementary materials.
 The best fit fault plane bisects the intermediate MT resistivity anomaly and EQ hypocenters cluster beneath the domed low-resistivity layer. These spatial relationships are consistent with a fracture-controlled geothermal system. PCA indicates that the EQ swarm is elongated to the SW. Emerging from the SW edge of the hypocenter swarm, there is a low Vp anomaly also extending SW to the coast in the direction of HWP. VP has been shown to decrease by roughly 10% when rock samples from high-temperature geothermal fields are heated from 50°C to 250°C [Kristinsdóttir et al., 2010]. When viewed in relation to the MT and EQ results, the Vp anomaly in the SW is consistent with a high-temperature geothermal reservoir. The geophysical data sets taken together produce a very compelling case for a fault fracture controlled high-temperature geothermal reservoir beneath SGH.
 The geophysically determined geothermal system lies at the intersection between the SW-oriented fault and what has been described as a WNW trending zone of weakness [Wadge and Isaacs, 1988]. The WNW zone is delineated by the alignment of Roche's Bluff, The Soufrière Hills volcanic centre, SGH, and GH. The system is situated at the intersection of two fault systems, where geothermal reservoirs often form [Faulds et al., 2011]. Fault intersections produce zones of enhanced permeability and also produce crustal weaknesses that can act as preferential magma pathways. The area beneath SGH has been the site of periodic seismic swarms that have been recorded in 1933–1937, 1966–1967, and 1992–1996; there was also a preinstrumental seismic crisis 1897–1898 [Aspinall et al., 1998] that was also associated with this area. These seismic crises are consistent with repeated magmatic injection episodes occurring beneath SGH as well as SHV. The 1992 injection episode culminated in the onset of eruption at SHV in 1995. A separate geothermal system beneath SGH is supported by this evidence of recent magmatic injection and is also consistent with significantly hotter shallow well temperatures obtained in areas closer to SGH than SHV as well as the discrete nature of the Vp anomaly in this area. The fault zone estimated using PCA extending to the SW is consistent with an upflow and outflow pathway that intersects HWP. The data are consistent with a geothermal system, associated with a fault intersection beneath SGH, with an upflow to the SW.
5 Prospectivity Index
 In this section, we develop a rational method to constrain a drilling program to intersect the most permeable and productive regions of the geothermal system.
 The geothermal power potential, or harvestable heat flux, of a particular zone in a geothermal aquifer will be linearly related to both the temperature and the permeability of that zone, because fluid temperature will be roughly proportional to its enthalpy. Similarly, through Darcy's law, flow rates will be linearly related to permeability. The conceptual model and geophysical data are used to determine a spatially varying prospectivity index, which is the product of measured proxies for temperature and permeability, where the highest values relate to the zones within the area which will be most productive in terms of thermal power.
 We assume that significant spatial variations in permeability are due to fault-related fracturing and that the subsurface temperature is inversely proportional to the value of the Vp anomaly. Mitchell and Faulkner  determine experimentally that fracture density F decreases exponentially with distance from the fault core, x, according to equation (1).
 Where F0 is fracture density near the fault core, and β is the distance over which fracture density falls to 1/e of its maximum value. We determine spatial variation in fracture density using equation (1) and the SW and WNW trending fault traces determined by PCA and structural alignments. The fracture density fields from each fault are summed linearly, yielding a maximum relative fracture density at the intersection zone. The fracture density field was converted to a relative permeability field using the power law relationship between permeability k and fracture density described by Mitchell and Faulkner  and demonstrated in equation (2).
 We use a value of 1.5 for the power law exponent γ, the same value determined empirically for faults within the Atacama fault system in Northern Chile. c is set to 1 since we are seeking to recover relative permeability variations. Relative permeability is highest at the fault intersection. The permeability map is included in the supplementary materials.
 A relative reservoir temperature proxy was attained by assuming a linear relationship between the negative Vp anomaly and temperature (Figure 3). We used Vp data at 1500 m depth for this analysis. This zone is beneath the interpreted clay cap and intersects the slowest region of the Vp anomaly. Relative permeability and temperature proxies were scaled to vary between 0 and 10. The two fields were then multiplied together to yield what we have called a prospectivity index (Figure 4).
 The prospectivity index represents a quantification of the conceptual model. The highest indices occur in zones expected to yield the most geothermal power as indicated by the geophysical data.
 The contours of the prospectivity index map (Figure 4) have contours which elongate in the directions of the modeled faults; however, the contours are stretched to the south west. This is due to the southwest offset of the slow Vp anomaly.
 The temperature-dependent Vp measurements of Kristinsdóttir et al.  should be applicable to Montserrat since these rocks contain hydrothermally altered minerals, such as smectite, chlorite, and epidote, which we would expect to be present in the Montserrat system. However, Vp variations may also be due to variation in porosity as well as degree of alteration [Han et al., 1986; Johnston and Christensen, 1997]. Increases in both alteration and porosity cause significant decreases in Vp velocity of the order observed in Montserrat. Observed hot spring and shallow borehole geothermal waters prove the existence of an active permeable geothermal system. Given that the flow of hot geothermal fluid through a reservoir causes hydrothermal alteration of the reservoir rocks [Browne, 1970], we argue that Vp velocity decreases due to both alteration and temperature are indicative of fluid flow through permeable zones of the reservoir. Hydrothermal alteration processes are on average volume neutral (P. R. L. Browne, personal communication), and fluid flow pathways tend to be kept open by the microseismicity, which initially form the permeable pathways [Zoback, 2007].
 Recent work by Hautmann et al.  shows significant density decreases at 1 km depth, consistent with the Vp anomalies observed. This observation does not definitively support any one interpretation since all three possible causes of Vp anomaly would also cause a decrease in density. However, it does strongly support the existence of a significant anomaly in rock properties in these zones.
 The other two anomalously slow zones to the NW and NE could be related to relict geothermal systems or active systems with no surface expression. These zones require further exploration.
 The permeability model used in this study based on the work of Mitchell and Faulkner  contains a number of arbitrary parameters. For example, the characteristic distance β is estimated based on the width of the intermediate-resistivity zone, and the exponent γ is given a value of 1.5, similar to that empirically derived from faults in Chile. Although these values are poorly constrained, variations would not change the fundamental shape of the prospectivity map but would serve to increase or decrease the contour spacing. This represents a fundamental limit to the approach followed in this study, and further field studies would be necessary to determine more appropriate values for these variables.
 As calculated, the prospectivity index is roughly linearly related to the harvestable heat flux. However, the efficiency of conversion of thermal power to electric power is related to the temperature of the resource through an efficiency factor which increases with the temperature of the resource—the theoretical limit is the Carnot efficiency. Taking this into account in calculating the index would shift higher index values toward the higher-temperature areas. For simplicity, we have not included this consideration in our index calculation.
 Due in part to the dense vegetation and lack of good exposure on the island, there is a lack of comprehensive structural geological data for the island which would be useful in further constraining conceptual models.
 Integration of magnetotelluric, seismic P wave tomography data, and EQ location data, in conjunction with the DEM of the island and existing geological data, was used to create a three-dimensional conceptual model of a geothermal system in western Montserrat. The data are consistent with repeated magmatic injections along a fault intersection in the region of SGH, which has created a heat source for a geothermal system in the area. The data also indicate that enhanced permeability due to the fault intersection hosts the geothermal fluid with an upflow beneath SGH and an outflow to the west near HWP. The geophysical data were used to create a prospectivity index map that was a product of a modeled permeability distribution and subsurface temperature proxy data derived from the Vp anomaly at 1500 m depth. The index is a spatially varying relative proxy for geothermal power potential. This information can be used in conjunction with logistical factors to determine the best areas to target for geothermal exploitation.
 We thank the government of Montserrat for permission to publish the results of this study. We are also grateful to Norman Ryan, the Montserrat Volcano Observatory staff, and its director Rod Stewart for significant logistical support. We thank Greg Scott (Caribbean Helicopters) and the British Royal Navy for flying us to difficult-to-access locations close to the volcano. We are grateful to Diana Roman for access to the seismic data used in this study, and we would also like to thank Mila Adams for informative discussions on the seismic properties of rocks. We also acknowledge Stephen Onacha for assisting with the processing of the magnetotelluric data. Andrew Newman and two anonymous reviewers are thanked for their comments which improved the manuscript.