Optimized multicomponent vs. classical geothermometry: Insights from modeling studies at the Dixie Valley geothermal area

Geothermics 51 (2014) 154–169

Contents lists available at ScienceDirect

Geothermics journal homepage: www.elsevier.com/locate/geothermics

Optimized multicomponent vs. classical geothermometry: Insights from modeling studies at the Dixie Valley geothermal area L. Peiffer a,∗ , C. Wanner a , N. Spycher a , E.L. Sonnenthal a , B.M. Kennedy a , J. Iovenitti b a b

Lawrence Berkeley National Laboratory, Berkeley, CA 94720, USA AltaRockEnergy Inc., Sausalito, CA 94965, USA

a r t i c l e

i n f o

Article history: Received 10 April 2013 Accepted 16 December 2013 Available online 15 January 2014 Keywords: Geothermometer Optimization Exploration Numerical modeling Mixing Geothermal

a b s t r a c t A new geothermometry approach is explored, incorporating multicomponent geothermometry coupled with numerical optimization to provide more confident estimates of geothermal reservoir temperatures when results of classical geothermometers are inconsistent. This approach is applied to geothermal well and spring waters from the Dixie Valley geothermal area (Nevada), to evaluate the influence of salt brines mixing and dilution of geothermal fluids on calculated temperatures. The main advantage of the optimized multicomponent method over classical geothermometers is its ability to quantify the extent of dilution and gas loss experienced by a geothermal fluid, and to optimize other poorly constrained or unknown parameters (such as Al and Mg concentrations), allowing the reconstruction of the deep reservoir fluid composition and therefore gaining confidence in reservoir temperatures estimations. Because the chemical evolution of deep geothermal fluids is a combination of multiple time-dependent processes that take place when these fluids ascend to the surface, reactive transport modeling is used to assess constraints on the application of solute geothermometers. Simulation results reveal that Al and Mg concentrations of ascending fluids are sensitive to mineral precipitation–dissolution affecting reservoir temperatures inferred with multicomponent geothermometry. In contrast, simulations show that the concentrations of major elements such as Na, K, and SiO2 are less sensitive to re-equilibration. Geothermometers based on these elements give reasonable reservoir temperatures in many cases, except when dilution or mixing with saline waters has taken place. Optimized multicomponent geothermometry yields more representative temperatures for such cases. Taking into account differences in estimated temperatures, and chemical compositions of the Dixie Valley thermal waters, a conceptual model of two main geothermal reservoirs is proposed. The first reservoir is located along the Stillwater range normal fault system and has an estimated temperature of 240–260 ◦ C. It covers the area corresponding to the geothermal field but could extend towards the south-west where deep temperatures of 200–225 ◦ C are estimated. The second reservoir has an estimated temperature of 175–190 ◦ C and extends from well 62-21 to northeastern Hyder, Lower Ranch, Fault Line, and Jersey springs. © 2014 Elsevier Ltd. All rights reserved.

1. Introduction Solute geothermometers have been applied during the last decades as an exploration tool for inferring geothermal reservoir temperatures. The most widely used are the ‘classical’ geothermometers, which are either based on an absolute concentration like dissolved SiO2 or on a ratio between dissolved elements such as Na–K, Na–K–Ca, or K–Mg (Fournier and Truesdell, 1973;

∗ Corresponding author. Present address: Instituto de Energías Renovables, Universidad Nacional Autónoma de México, Temixco, Morelos 62580, Mexico. Tel.: +52 5556229726. E-mail addresses: [email protected] (L. Peiffer), [email protected] (N. Spycher). 0375-6505/$ – see front matter © 2014 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.geothermics.2013.12.002

Fournier, 1979; Fournier and Potter, 1982; Giggenbach, 1988). Another approach presented by Reed and Spycher (1984) involves the computing of saturation indices (log(Q/K)) of potential reservoir minerals over a range of temperatures, given a full thermal water analysis, then estimating the reservoir temperature by the clustering of saturation indices near zero. This multicomponent geothermometry approach was recently formulated into a computer program (GeoT) that automatically estimates reservoir temperatures (Spycher et al., 2011, 2014). As a stand-alone program, GeoT can be coupled with existing parameter estimation software such as iTOUGH2 (Finsterle and Zhang, 2011) to allow numerical optimization of unknown or poorly constrained input parameters that may adversely affect temperature estimations. This new integrated multicomponent approach is tested here in context with results of numerical simulations, fluid analyses and

L. Peiffer et al. / Geothermics 51 (2014) 154–169

155

Fig. 1. (A) Aerial view of the Dixie Valley geothermal area and surroundings with locations of wells and springs (Google Earth image). Wells: s33 – Section 33, s7 – Section 7, s18 – Section 18. Springs: D – Dixie, B – Big Horn, S – Sou, H – Hyder, FL – Fault Line, LR – Lower Ranch, M – McCoy, J – Jersey. All waters were sampled by Goff et al. (2002). (B) Structural map of the Dixie Valley (modified from Iovenitti et al., 2012). Note that Jersey springs are off-map. The dashed white line on both maps corresponds to the cross-section shown in Fig. 2.

mineralogical data from the Dixie Valley geothermal system, and classical geothermometry. All solute geothermometry methods present advantages and disadvantages. All approaches rely on the assumption that the deep reservoir fluid is unmodified by physical (e.g., mixing, dilution, and boiling) or chemical processes during ascent to the surface and that the chemical composition sampled at the surface reflects chemical equilibrium (or near-equilibrium) with reservoir minerals over some narrow range of temperature. When re-equilibration has taken place upon cooling and/or boiling, geothermometers typically miscalculate reservoir temperatures. A detailed discussion on the application range of classical solute geothermometers was presented by Giggenbach (1988). The Na/K geothermometer, based on the equilibrium of geothermal fluids with Na and K feldspars, has been shown to be the least affected by re-equilibration processes due to the slow response of the Na/K ratio to re-equilibration. Therefore, this geothermometer is typically assumed to reflect the deepest equilibration temperature. On the opposite side, the K/Mg geothermometer, based on the fluid equilibration with muscovite, clinochlore and K-feldspar, re-equilibrates much faster and has been used to infer the temperature of last equilibrium. The Na–K–Ca geothermometer is also widely used, however it is known to be sensitive to calcite precipitation and thus to the CO2 content of geothermal fluids. The mineral saturation index approach presented by Reed and Spycher (1984) is more integral because it is truly multicomponent and based on any number of minerals. However, the fact that this method relies on complete analyses of thermal waters, and requires numerical computations, can be viewed as a significant practical disadvantage compared to classical geothermometry. Also, by searching for the convergence (near zero) of saturation indices of multiple minerals, the method assumes that these minerals are all at or near equilibrium with the deep fluid, which may not always be the case for the all minerals considered. This multicomponent approach is also known to be sensitive to Al concentrations (Pang and Reed, 1998), and our own efforts have shown that this method can be quite sensitive to Mg

concentrations as well, as discussed later in this paper. This can present difficulties because Al analyses are often missing in geochemical studies, or are easily affected by errors because of typically low Al concentrations in geothermal waters. In this work, the effect of reactive processes such as mineral precipitation and dissolution affecting deep geothermal fluids as they ascend to the ground surface is investigated, and specific cases of reliable and unreliable geothermometry results are presented. This is achieved by running reactive transport simulations and applying both classical and multicomponent geothermometry to different groups of thermal waters (wells and springs) from Dixie Valley, Nevada. Optimized multicomponent geothermometry computations (Spycher et al., 2011, 2014) are applied to reconstitute the deep original geothermal fluid composition in order to estimate the deep reservoir temperature. Groups of waters are processed simultaneously to better constrain the optimization process, under the assumption that waters in these groups originate from a common location. The grouping of these waters is established using standard hydrogeochemical analyses such as elemental and isotopic correlation plots. The effect of Al and Mg concentrations on chemical geothermometry is also investigated, as well as the impact of dilution and mixing with saline brines. This study is augmented by a more comprehensive modeling investigation of Dixie Valley presented by Wanner et al. (2013, 2014), who developed a two-dimensional reactive transport model across the valley to investigate deep fluid and heat flow patterns, as well as reactive processes affecting solute geothermometry. Dixie Valley was selected as an area of study not only because of the large number of water chemistry analyses available for this geothermal area (Goff et al., 2002), but also because it includes a diversity of waters with different chemical compositions, and for which classical geothermometers yield a large range of reservoir temperatures, especially when applied to thermal springs. This area has also been the subject of numerous investigations, including a large synthesis report by Blackwell et al. (2007a), summarized in

156

L. Peiffer et al. / Geothermics 51 (2014) 154–169

Fig. 2. Geologic cross-section of the Dixie Valley area (Iovenitti et al., 2012). Temperature isotherms are also represented (◦ C), from data measured in geothermal wells (static temperature measurements). Abbreviations correspond to the different geological units: Tbf, tertiary sediments; Tmb, miocene basalts; Tv, oligocene welded tuffs; Kgr, cretaceous granodiorite intrusives; Jz, jurassic lopolith; Tr, triassic marine metasedimentary (see text).

Blackwell et al. (2007b), from which much of our data has been obtained. Before presenting geothermometry analyses for Dixie Valley, we summarize key data from this area and re-examine water compositions and chemical trends to gain further insight on this Basin and Range geothermal system. Finally, an interpretation of reservoir temperature distributions for the Dixie Valley geothermal area is proposed. 2. Geological settings of the Dixie Valley geothermal system A considerable number of studies were done on Dixie Valley providing numerous and diverse site information (seismic reflection, gravimetry, magneto-tellurics, thermal gradients, wells cuttings analyses, hydrogeochemistry; Blackwell et al., 2007a,b and references therein). Dixie Valley (150 km east of Reno, Nevada) is a 120 km long by 20 km wide north-northeast trending extensional basin, surrounded to the west by the Stillwater Range and East Range and to the east by the Clan Alpine-Augusta Mountains and Tobin Range (Fig. 1A). A 63 MWe geothermal power plant has been operating at Dixie Valley since 1986. The plant is located in the western part of the valley along the Stillwater Range fault system. The Stillwater fault system is a high-angle normal fault system consisting of a range-front and piedmont fault based on well, gravity, and magnetic data (Blackwell et al., 2007a,b). Recent integrated geological cross-sections focused on the central and western part of the valley, and jointly inverted gravity and magnetic modeling, support the existence of two major normal faults system along the Stillwater Range reported by Blackwell et al. (2007a,b). These faults represent potential structural controls of the shallow thermal anomalies along both sides of the Stillwater Range and the producing geothermal reservoir (Iovenitti et al., 2012). A series of normal faults are also observed throughout the valley (Figs. 1B and 2). Well cuttings reveal up to 2000 m of Tertiary basin-filling sediments composed of colluvial and alluvial gravels, sands, eolian silts, and lacustrine playa silts and clays (Tbf, Fig. 2). Deeper rocks units consist of Triassic marine metasedimentary rocks (mainly carbonates) (Tr) overlain by the Jurassic Humboldt Lopolith (Jz; Speed, 1976). The lopolith is composed of oceanic crustal rocks of gabbroic, dioritic and basaltic composition (Waibel, 1987; Lutz et al., 1997). Cretaceous granodiorite plutons (Kgr) intrude the entire Mesozoic sequence and are found in some wells at the footwall

of the Stillwater fault. Sequences of Oligocene silicic welded tuffs (Tv) are sometimes observed as thin layers (<55 m) in some wells above the lopolith. Miocene basaltic lava flows (Tmb) overlay earlier stratigraphy and appear in wells at depths of 1800–2300 m beneath the younger Tertiary basin-filling sediments. Two producing host rocks are exploited by the geothermal power plant. Most wells produce from a reservoir located in the Jurassic Humboldt Lopolith at a depth of 2830–3330 m where most of the productive fractures are localized (Waibel, 1987; Lutz et al., 1997). The reservoir is localized close to the Piedmont fault of the Stillwater fault system and has a fracture-dominated permeability (Williams et al., 1997). Some wells (Section 18) are shallower and produce from the upper Miocene basalt. Bottom hole temperatures (BHT) measured inside the Lopolith range from 240 to 250 ◦ C while lower temperatures of 220–230 ◦ C are reported inside the basalt. There is no significant public domain data on the presence or absence of a geothermal reservoir above the Miocene basalts (Blackwell et al., 2007a,b). Most geological units are hydrothermally altered. On the basis of XRD analyses of well cuttings, Lutz et al. (1997, 1998) identified six grades of alteration inside the Jurassic reservoir with post-Oligocene to present ages. From oldest to youngest the following assemblages were observed: (1) epidote–chlorite–calcite veins, (2) illite, (3) wairakite–quartz–calcite–potassium feldspar–epidote veins, (4) mixed-layers of illite and smectite, and quartz–calcite veins, (5) chalcedony–quartz–dolomite–calcite–chlorite–smectite and barite–hematite veins, and (6) quartz–calcite veins. Quartz and calcite from the late stage precipitate from the current geothermal fluid, as revealed by fluid inclusions showing homogenization temperatures of 200–270 ◦ C and salinities of 0.0–0.5 wt%. Bruton et al. (1997) characterized by XRD the scale deposits from production wells and identified amorphous silica as the major phase, and some minor amounts of quartz, calcite, magnetite, goethite, and unidentified clays. The Miocene basalt, representing the youngest volcanic activity inside the valley is too old (K–Ar age of ∼8.5 Ma; Waibel, 1987) to explain the high thermal regime of the area (Kennedy and van Soest, 2006). Instead, the geothermal fluid production is mostly associated with deep fluid circulation along the Stillwater fault system (Sass, 1995; McKenna and Blackwell, 2004). Kennedy and van Soest (2006) affirm that the helium isotopic composition of the geothermal fluid (0.70–0.76 Ra) corresponds to a 7.5% contribution of mantle–origin helium, and that the permeability induced by the Stillwater fault system allows deep magmatic fluids (from the crust–mantle transition zone) to mix with reservoir fluids. The geothermal field is divided (areally) into three main sections from north to south, Sections 33, 7 and 18 respectively (Fig. 1A). Section 7 and Section 33 wells are used as producers, and Section 18 wells as injectors. Although the average temperature gradient of the Dixie Valley area is about 63 ◦ C/km, it can reach values that are much greater than 100 ◦ C/km (Blackwell et al., 2007a). In particular, temperature gradients on the order of 100 ◦ C/km are inferred for the actual geothermal field along the Stillwater fault system because reservoir temperatures in excess of 250 ◦ C were measured at 3 km depth in Sections 7, 33 and 18 wells. A high temperature gradient is also manifested by near surface groundwater temperatures that can locally reach temperatures greater than 100 ◦ C (Iovenitti et al., 2012). Several groups of hot springs (T = 20–84 ◦ C; Goff et al., 2002) appear along the Stillwater range-front fault but also to the northeast in the valley: Dixie, Big Horn, Sou, Hyder, Fault Line, Lower Ranch, McCoy, Jersey. In the mountain ranges surrounding the valley, several cold springs and streams are also present (Fig. 1A). However, run-off water from the ranges most likely does not infiltrate very deep in the valley because rates of evapotranspiration are high. Harrill and Hines (1995) estimated that the recharge

L. Peiffer et al. / Geothermics 51 (2014) 154–169

157

by precipitation of the entire valley (2.8 × 107 m3 /year) is roughly balanced by the discharge by evapotranspiration (2.5 × 107 to 3.8 × 107 m3 /year). Nowadays, inside the valley, the Humboldt Salt Marsh is the only natural stagnant water body (Fig. 1A). A ‘Dead Zone Area’ (a local zone of plant die-off), extends from the northwestern part of the geothermal field towards the Stillwater range-front fault, and is characterized by fumarolic activity and steaming ground (Bergfeld et al., 2001). The whole valley is also characterized by shallow aquifers that are intercepted by shallow wells and used for water supply.

3. Water chemical composition and trends Nimz et al. (1999) showed that waters from the Stillwater Range have a higher Cl/HCO3 ratio than waters from the Clan Alpine and Augusta Ranges, while hot springs and cold waters from the valley are not systematically associated with one of the two groups. Geothermal well waters belong to the enriched-Cl type of fluid but show low Ca–Mg concentrations typical of high temperature fluids. Goff et al. (2002) presented an extensive set of geochemical data (major and trace elements, isotopes) of waters from the valley and surrounding ranges. We present a short interpretation of their data to distinguish different water chemistry groups that can be used for geothermometry investigations. The waters reported by Goff et al. (2002) were sampled from the production wells, pre-production wells (sampled in 1986 before re-injection), other geothermal and on-site wells (referred to as ‘ogo’ wells throughout the study, following Goff et al., 2002), valley thermal springs, and cold springs or wells from the valley and ranges. ‘ogo’ wells are non-productive and generally shallow (<300 m). Most of these shallow wells are used for domestic water supply and water reinjection. Wells 45-14 and 62-21 (Fig. 1A) were drilled as geothermal wildcat wells. These are deep wells (2440 m and 3810 m, respectively) but they are poor producers presumably because of low permeability. Plotting Na vs. Cl concentrations defines three main trends amongst all the waters (Fig. 3). North-east valley springs Hyder, Jersey, Fault Line, Lower Ranch and well 62-21 are characterized by low Cl/Na ratio. Pre-production geothermal fluids, most ‘ogo’ wells, Dixie Springs, McCoy springs and some cold waters define an intermediate Cl/Na trend. Well 45-14, located 15 km southeast from the production wells, is the only ‘ogo’ well with a significantly higher Cl content (481 mg/l) than the pre-production fluids (320–428 mg/l). Sou hot springs lie in between the low and intermediate Cl/Na trends. The third trend is defined by Big Horn springs and some cold springs and wells. It corresponds also to the pure halite dissolution trend (Cl/Na molar ratio of 1). The production well waters show a shift towards slightly higher Cl/Na compared to pre-production wells, probably due to the fact that injection fluids are a mix between evaporated brines and superficial cold well water, as suggested by Kennedy et al. (1999). The occurrence of evaporitic salts in the shallow subsurface is probably one important source of Cl. It is known that the Humboldt Salt Marsh area was exploited for salt deposits in the 19th century. The salt deposits were found to a depth of more than 30 m and composed of 97% NaCl, 2% Na2 SO4 , 1% Na2 CO3 . Brines within the salt had the following composition: 27% NaCl, 5% Na2 SO4 , 4% Na2 CO3 , 0.19–0.75% K2 O (Phalen, 1919; Vanderburg, 1940). The salt brine has the same Cl/Na ratio as the marsh brines (dashed line in Fig. 3). Salt leaching or mixing of the deep reservoir fluid with superficial saline brines could explain why some spring and shallow well waters have Cl/Na molar ratio close to 1. In this case, classical Na–K based geothermometers yield low temperature estimates.

Fig. 3. (A) Na vs. Cl concentration plot for Dixie Valley waters. (B through E) ı18 O–ıD–Cl relationships for Dixie Valley waters. DMWL, Dixie Valley Meteoric Water Line; WMWL, World Meteoric Water Line. Samples surrounded by the ellipse (except Big Horn) were sampled in the surrounding ranges. Data are from Goff et al. (2002). Abbreviations for wells and springs are the same as in Fig. 1.

Looking more carefully at pre-production well data, the concentrations of Na and Cl increase from Sections 33, 7 and 18, and geographically from NE to SW (black arrow in Fig. 3A). Well 45-14 is also part of this trend. Furthermore, lower measured downhole temperatures were reported for well 65-18 (225 ◦ C) and well 45-14 (196 ◦ C), compared to temperatures between 240 and 250 ◦ C for Sections 7 and 33 pre-production wells (Benoit, 1989; Blackwell et al., 2007a). Moreover, the increase in Cl and Na is clearly enhanced southward, towards the marsh. These trends are difficult to observe in the production well waters because these fluids are mainly flashed brines mixed with waters from shallow wells (Kennedy et al., 1999). With the exception of geothermal wells, ‘ogo’ wells and some hot springs, most waters from the valley and surrounding ranges present ␦18 O–␦D compositions similar to the local Dixie meteoric

158

L. Peiffer et al. / Geothermics 51 (2014) 154–169

water line values (DMWL; Fig. 3B). The cold waters from the surrounding ranges generally show less negative ␦D values than the valley waters. According to Nimz et al. (1999), this difference in ␦D indicates that the recharge in the valley occurred during a colder period than for the range waters. Pre-production well waters and their corresponding pre-flashed composition are also plotted (empty and black diamonds). The pre-flashed values show a shift in ␦18 O of ∼+2‰ compared with cold valley waters, probably reflecting water–rock interaction at high temperature (Giggenbach and Stewart, 1982). Production well waters show typical ␦D–␦18 O shift due to boiling, while the ‘ogo’ wells isotopic compositions are dispersed between the pre-production and current production wells and the meteoric values of the valley waters. Amongst the north-east valley springs, Hyder springs show the most significant oxygen shift (>+1‰), similar to well 62-21. Well 45-14 is characterized by a ␦18 O shift of +1.2‰. Apart from the pre-production wells, pre-flashed isotopic compositions could not be calculated due to the lack of historical phase separation data. Kennedy et al. (1999) proposed the existence of a compartmentalized reservoir or adjacent isolated reservoirs with different temperature regimes and slightly different water compositions. This hypothesis is confirmed by the ␦18 O and ␦D plots vs Cl (Fig. 3C and D). No clear trend of mixing or boiling can link the Cl and oxygen-18 enriched waters (pre- and production wells, well 4514, Dixie, Big Horn springs and some cold waters) to the other Dixie Valley waters. The only clear trend is that described by the pre-production and production wells due to boiling. Furthermore, north-east springs Hyder, Jersey, Fault Line, Lower Ranch probably originate from a common reservoir. Dilution by meteoric fluid of the original deep fluid might cause the small spread of values described by those springs. Well 62-21 and Sou hot springs also seem to be related to the same source, while Dixie and McCoy springs are characterized by Cl enrichment. Although pre-production wells are characterized by increased Cl concentration towards the southwest (from Section 33 to Sections 7 and 18), they have very similar pre-flash ␦D (−133 to −131‰) and ␦18 O values (−15.0 to 14.6%), reflecting most probably a common origin, but clearly different from the northeast springs. Using carbon and chloride isotope data (14 C and 36 Cl/Cl), Nimz et al. (1999) calculated ages for Clan Alpine waters of less than 50 years, and Stillwater Range water ages greater than 1000 years. Geothermal waters are estimated to be 12–14 ky old (Late Pleistocene). At that time, a pluvial lake covered most of the valley (Thompson and Burke, 1973). One hypothesis is that the lake water infiltrated along the Stillwater fault system and experienced water–rock interaction at elevated temperature (Nimz et al., 1999). Consequently, this infiltrated water could represent the source of the modern geothermal fluids. The marsh salt deposits are believed to be the remnants of the evaporated Pleistocene lake. The difference in age between the range and valley waters, as well as their ␦18 O–␦D composition, appears to rule out the possibility that modern water from the ranges could recharge the deep geothermal reservoir of the valley (Nimz et al., 1999). Dixie Lake paleo-shorelines show that the lake did not extend to the eastern part of the valley (Reheis, 1999), and this could explain the low Cl/Na trend of the north-east springs.

4. Optimized multicomponent geothermometry approach Multicomponent geothermometry, based on computed mineral saturation indices, relies on complete chemical data sets, including Al and Mg. The choice of a relevant mineral assemblage and accurate chemical analyses are keys to the method’s success (Reed and Spycher, 1984; Pang and Reed, 1998). To our knowledge, only two

studies previously applied multicomponent geothermometry to Dixie Valley fluids. Bruton et al. (1997), in their preliminary study of Dixie Valley scale formation and production wells water chemistry, computed mineral saturation indices as a function of temperature for water samples from a production well (76-7). These authors used the reported total dissolved (0.45 ␮m filtered) Al concentration and considered five minerals: quartz, K-feldspar, wairakite, calcite, and clinozoisite (epidote). They obtained a large spread in temperatures (220–280 ◦ C) considering the four first minerals, while clinozoisite was computed to remain above saturation levels over the entire temperature range. The consideration of wairakite as a potential reservoir mineral may be in question because this mineral is rarely observed in borehole cuttings (Lutz, unpublished data) and it is not considered to be deposited by the actual geothermal fluids (Lutz et al., 1998). Shevenell and De Rocher (2005) reported two multicomponent geothermometry results using the composition of fluids from well 28-33 and a list of twelve common alteration minerals. They first used reported total dissolved (0.45 ␮m filtered) Al concentrations (Lisa Shevenell, personal communication) and obtained a low temperature estimate of 200 ◦ C. A second temperature estimate was higher (233 ◦ C) and was obtained by constraining the Al concentration to reflect equilibrium with kaolinite. The same authors mentioned a poor clustering of the saturation indices curves at this temperature and concluded that the multicomponent geothermometry method may be more suitable for lower temperature reservoirs. These earlier studies, and our own evaluations presented below, clearly show that the selection of relevant minerals is critical for reliable temperature estimations, and that the Al and Mg concentrations also clearly affect temperature estimations at Dixie Valley. We discuss below the further application of multicomponent geothermometry to Dixie Valley waters using the newly developed code GeoT (Spycher et al., 2011, 2014), and the integration of this method with numerical optimization to refine temperature predictions by allowing the estimation of poorly constrained Al and Mg concentrations, as well as other parameters affecting the geothermometry calculations.

4.1. Mineral assemblage and thermodynamic data For the case of Dixie Valley, a mineral assemblage was selected on the basis of XRD analyses of cuttings from the drilling of geothermal wells (well 73B-7; Lutz, unpublished data; Table 1). Those XRD data were obtained on bulk rock samples. Quartz, calcite, microcline and albite are present in the entire stratigraphic column; clinochlore (chlorite) in the lopolith and basalt layers, laumontite only in the basaltic layer, and montmorillonite and Na–clinoptinotile (zeolite) only in the basin-filling sediments horizon. Pyrite and goethite are the main iron minerals present throughout the column. Epidote minerals were only observed in the cuttings from a few boreholes in the gabbroic horizon (<2 wt%) and were not considered in the simulations. The thermodynamic database SOLTHERM.H06 (Reed and Palandri, 2006) was used, which incorporates data for minerals primarily from Holland and Powell (1998), and for aqueous species primarily from SUPCRT92 (Johnson et al., 1992). This database was developed specifically for elevated temperature applications. It has been tested with simulations of numerous types of hydrothermal systems, and continues to be refined. As further discussed by Spycher et al. (2014), thermodynamic data are critical inputs for multicomponent geothermometry, and have a direct effect on temperature predictions.

L. Peiffer et al. / Geothermics 51 (2014) 154–169 Table 1 Main rock formations at Dixie Valley and corresponding initial mineral volume fractions (Lutz, unpublished data) specified for reactive transport simulations. Rock formation

Mineral

Volume fraction

Lopolith (Jz)

Calcite Quartz Albite Microcline Clinochlore Pyrite

0.11 0.47 0.38 0.01 0.02 0.01

Basalt (Tmb)

Calcite Quartz Albite Microcline Clinochlore Laumontite Goethite

0.12 0.14 0.49 0.14 0.03 0.01 0.07

Basin-filling sediments (Tbf)

Calcite Quartz Chalcedony Amorphous silica Albite Microcline Clinoptinolite–Na Montmorillonite–Ca Goethite

0.12 0.17 0.00 0.00 0.32 0.24 0.01 0.11 0.03

4.2. Initial geothermometry calculations – fluid analyses issues It has long been known that dissolved Al concentrations can be affected by the formation of colloids (e.g., Barnes, 1975) and, in the case of geothermal systems, by the formation of colloidal and/or solid Al phases upon cooling. Furthermore, the analyses of Al are often not reported in geochemical studies. Pang and Reed (1998) recognized this problem and proposed to constrain the multicomponent geothermometry approach by computing the Al activity assuming equilibrium of the fluid with a selected Al-bearing solid phase at all temperatures (i.e., their “Fix-Al” method). The “Fix-Al” method has shown to be quite useful, however its results can be quite sensitive to the choice of the mineral selected to fix Al activity. Another approach proposed further below is to constrain dissolved Al concentrations by numerical optimization. For the case of Dixie Valley, Goff et al. (2002) were aware of the challenge regarding analyses of Al concentrations. For this reason these authors reported both “total” and “dissolved” (ionized) Al analyses in filtered samples (0.45 and 0.2 ␮m, respectively). For the dissolved Al analysis, they used the oxine-MIBK method from Barnes (1975), which allows the separation of dissolved monomeric Al species (e.g., Al3+ , AlOH2+ , Al(OH)2 + , and Al(OH)4 − ) that are considered to participate in chemical equilibria at the time the water is sampled, from the metastable and slowly reacting macro-ions and colloidal particles. This technique allows an efficient removal of the colloids that may not be completely removed by filtration. The availability of these two types of Al analyses provided an opportunity for us to further investigate the sensitivity of the temperatures estimated with GeoT to Al concentrations, and to compare modelpredicted Al concentrations to actual measurements. The application of GeoT to the geothermal Section 7 production well 73B-7, using the 0.2 ␮m-filtered dissolved Al concentration (0.063 mg/l) and the full list of minerals selected for this system, yields a temperature significantly lower than measured downhole temperatures (188 ◦ C vs. 240–250 ◦ C) (Fig. 4A). Note however that a large spread of temperatures (92 ◦ C) is indicated by the log(Q/K) curves of minerals crossing the equilibrium point (zero log(Q/K) values). This suggests that these minerals should not all be considered together when making temperature determinations, because they may not all be near equilibrium with the reservoir fluid. Clinochlore was observed in cuttings from boreholes drilled into the reservoir,

159

and is used here as the Mg-bearing phase. Because Mg was not detected in this water, the input Mg concentration was taken as the detection limit (0.01 mg/l). In doing so, the clinochlore saturation indices yield an equilibration temperature (210 ◦ C) that is lower than the reservoir temperature, because the detection limit overestimates the actual Mg concentration at depth. By lowering the input Mg concentration by a factor of ten (to 0.001 mg/l), the clinochlore equilibrium temperature is shifted to 264 ◦ C (not shown), which illustrates clearly the sensitivity of the computed temperature to Mg concentrations. Considering only the main silicate minerals known to be present in the reservoir (in the lopolith: quartz, albite, microcline and clinochlore) yields a temperature of 208 ◦ C (Fig. 4B), but this estimate is still lower than the measured downhole temperatures, and the spread in temperatures computed with these four minerals is still large (64 ◦ C). In contrast, when using the total Al (0.45 ␮m filtered) concentration (1.04 mg/l), the same fluid and the full mineral assemblage, a temperature of 264 ◦ C is obtained (Fig. 4C), closer to measured downhole values. However, in this case, the clustering of the curves, as a whole, is poor, with a range of temperatures covering 140 ◦ C. A higher temperature of 284 ◦ C is obtained if only the main silicate minerals present in the lopolith are considered. However, in this case, the temperature estimate is still associated with a large uncertainty (spreading 112 ◦ C) because of the Mg concentration issue affecting the saturation indices of clinochlore (Fig. 4D). By forcing equilibrium with albite and clinochlore to constrain, respectively, the Al and Mg concentrations at all temperatures, the temperature estimated with GeoT using the four main minerals in the lopolith (268 ◦ C) is above measured downhole values. The same procedure using the full list of minerals yields a lower temperature (184 ◦ C) (Fig. 4E and F). It should also be noted that the forced-equilibration method can be prone to convergence problems. In addition, constraining the activity of certain elements by forcing equilibrium with given minerals at all temperatures tends to dampen the slopes of the log(Q/K) curves with temperature, which then can affect the GeoT temperature estimation procedure. For these reasons, we propose below another method to constrain dissolved Al and Mg concentrations by numerical optimization. It should be noted that in the present case, classical geothermometers seem to yield reasonable temperatures (Na–KFournier79 253 ◦ C, quartzFournier&Potter82 240 ◦ C) once waters have been corrected for gas losses (following the method of Reed and Spycher (1984) implemented into GeoT). However, as discussed later in this paper, several Dixie Valley spring compositions seem to reflect salt leaching and/or mixing with shallow saline waters (suggested by molar Na/Cl ratio values near 1), yielding Na–K ratios which are no longer representative of deep fluids and cannot be used directly for geothermometry. The dilution of spring waters also typically affects SiO2 -based geothermometers. In such cases, multicomponent geothermometry, coupled with numerical optimization to constrain unknown or poorly known input parameters, provides a means to gain confidence in estimated reservoir temperatures. Furthermore, because this method improves on the reconstruction of deep fluids, applying classical geothermometers with the reconstructed fluids further increases confidence when results of both classical and multicomponent geothermometry converge. The optimized multicomponent geothermometry procedure is discussed below, and results are presented later in this paper. 4.3. Numerical optimization procedure In this study, GeoT was coupled with the parameter estimation software iTOUGH2 (Finsterle and Zhang, 2011), as proposed by Spycher et al. (2014). Earlier tests using PEST (Doherty, 2008) were also successful. In short, these stand-alone optimization codes can

160

L. Peiffer et al. / Geothermics 51 (2014) 154–169

4

3

A

1 0 -1 calcite microcline pyrite clinoptinolite-Na montmorillonite-Ca

-3

-4

C

0 -1 -2

quartz albite-lo clinochlore laumontite goethite

0 -1

0 -1

quartz

-3

4

3

2

2

184°C

-1 -2 -3

-4

calcite microcline pyrite clinoptinolite-Na montmorillonite-Ca

quartz albite-lo clinochlore laumontite goethite

albite-lo

clinochlore

100 120 140 160 180 200 220 240 260 280 300 Temperature (°C)

Fix-Al-Mg

0

microcline

-4

Log(Q/K)

Log(Q/K)

284°C

1

3

1

total <0.45 µm Al

-2

quartz albite-lo clinochlore laumontite goethite

100 120 140 160 180 200 220 240 260 280 300 Temperature (°C)

E

clinochlore

2 Log(Q/K)

Log(Q/K)

1

4

D

3

264°C

-4

albite-lo

100 120 140 160 180 200 220 240 260 280 300 Temperature (°C) 4

calcite microcline pyrite clinoptinolite-Na montmorillonite-Ca

microcline

-4

total <0.45 µm Al

2

-3

quartz

-3

3

-2

208°C

1

100 120 140 160 180 200 220 240 260 280 300 Temperature (°C) 4

dissolved <0.2 µm Al

2

188°C

-2

B

3

Log(Q/K)

Log(Q/K)

2

4

dissolved <0.2 µm Al

F

Fix-Al-Mg

268°C

1 0 -1 -2 -3

quartz

microcline

albite-lo

clinochlore

-4 100 120 140 160 180 200 220 240 260 280 300 Temperature (°C)

100 120 140 160 180 200 220 240 260 280 300 Temperature (°C)

Fig. 4. Mineral saturation indices (log(Q/K)) computed with GeoT for water sampled in well 73B-7. The estimated temperatures are given by the minimum median of absolute log(Q/K) values. See text for discussion of cases A–B–C–D–E–F.

run any model multiple times and determine response surfaces for variations of any input parameters, then estimate the values of these parameters that minimize the model departure from given data (i.e., that minimize an objective function). Reasonably good results were obtained by using one of the following minimization algorithms implemented into iTOUGH2: grid search, Levenberg–Marquardt, and simplex methods (Finsterle and Zhang, 2011). The minimization was applied directly to: (1) the median of the absolute values of the saturation indices of all minerals, for each water considered, (2) the spread of temperatures given by the equilibrium point (log(Q/K) = 0) of each mineral individually, and (3) when processing multiple water analyses simultaneously, the standard deviation associated with the average of the different individual temperature estimates. It should be noted that further

investigations on the use of efficient and reliable objective functions, minimizing procedures and search functions for applications to multicomponent geothermometry are ongoing and not within the scope of the present study. The optimized multicomponent geothermometry approach allows estimating unknown GeoT input parameters, such as the amount of water diluting the deep geothermal fluid on its way to the surface (‘cfact’), the percentage of gas lost during boiling (steam fraction ‘sf’), as well as the concentrations of any dissolved element for which analyses are missing or erroneous (e.g., Al and Mg). Several waters can be processed simultaneously to better constrain the optimization process, as long as these waters originate from a common reservoir (otherwise unreliable results are be obtained).

cfact

– – – 1.6 – – 0.06 0.07

Mg Al

– – 0.40 0.12 0 0 4 11 250 70 250 250 437 229 308 296 272 32 217 154

Optimized parameters Tstdev TgeoT TK–Mg TNa–K–Ca

245 75 245 149

(G)

TNa–K

231 54 231 129 238 74 237 203 980 791 1000 683 130 130 130 86 352 352 352 2015 1.17 2.64 0.20 0.19 0.32 0.03 5.74E−07 3.43E−06 0.07 0.02 1.67 1.14 0.12 97 14 21

Fe

Cl

SO4

HCO3

Tquartz

TNa–K

(F)

66 2 66 44

(mol m−2 s−1 ), Ea is the activation energy (kJ/mol), Q is the ion activity product and K is the equilibrium constant of the reaction. Reactive rate constants for each mineral were taken from a compilation by Palandri and Kharaka (2004). For dissolution, initial (base-case) specific reactive surface areas (A) of 324 cm2 /g were used for all minerals, except for montmorillonite, for which a higher surface area of 6824 cm2 /g was specified. These values, taken from Dobson et al. (2003), were estimated for volcanic rocks by calculating the geometric surface area of spherical and flat grains about 60 ␮m in size. For precipitation, the surface areas of minerals already present in the rock (see Table 1 for volume fraction) were assumed to be the same as the input surface areas for dissolution; for new secondary minerals (e.g. chalcedony and amorphous silica), surface areas were initially computed assuming a small mineral seed (of volume fraction ∼ 10−8 ). The same thermodynamic database (SOLTHERM.H06, Reed and Palandri, 2006) was used as for the GeoT computations. A deep reservoir water (Table 2) was reconstructed by first considering an average of analyses of production well waters and

565 349 565 1530

2

412 26 409 272

2

face area (m2 kg−1 s−1 ), k25 is the reaction rate constant at 25 ◦ C H O

8.00 6.32 6.95 7.03

s−1 ), A is the reactive surwhere r is the reaction rate (mol kg−1 H O

250 70 70 70

(1)

A B C D



1 1 2 3

Q K

Al

· 1−

Mg

 

Ca

T

1 298.15

K

R

−

Na

E 1 a

SiO2

r = A · k25

pH

The numerical mesh consists of a 3500 m deep, vertical 1D column that was discretized into 5 m grid blocks. To represent flow in a fracture in contact with wall-rock minerals, the volume of each grid block was divided into half void space and half matrix minerals. A constant flow rate was maintained in the fracture by specifying a fixed injection rate at the bottom of the column, which was set to a high permeability (i.e., no permeability constraint). A constant and linear temperature gradient was assumed between the bottom of the column (250 ◦ C) and the ground surface (70 ◦ C). A temperature of 70 ◦ C corresponds to the upper range of observed spring water temperatures at Dixie Valley (Goff et al., 2002). The column was divided into the three main geological formations observed during drilling (Fig. 2): (1) lopolith rocks (Jz) from 3500 m to 2710 m below ground surface (bgs); (2) basaltic rocks (Tmb) from 2710 m to 2200 m bgs; and (3) basin-filling sediments (Tbf) from 2200 m bgs to the surface. The thin silicic welded tuff layer (Tv) (Fig. 2) was not considered because it is not observed in most wells. Mineralogy and respective volume fractions were assigned according to XRD analyses of cuttings obtained when drilling the geothermal wells (Lutz, unpublished data; Table 1). Chalcedony and amorphous silica were specified as potentially forming secondary phases in the upper layers of the modeled column. All minerals were set to react under kinetic constraints, except for calcite and goethite which were set to react at equilibrium. Precipitation and dissolution reactions under kinetic constraints were computed using the following equation, derived from transition state theory (Lasaga, 1984):

T (◦ C)

5.1. Model setup

#

Reactive transport simulations were performed to examine the chemical evolution of a deep geothermal fluid as it ascends to the surface, and to assess constraints on the application of solute geothermometers, including the optimized multicomponent geothermometry approach discussed earlier. This was achieved using the TOUGHREACT V2.0 reactive transport code (Xu et al., 2011) with a simplified one-dimensional (1D) model domain incorporating geochemical inputs relevant to Dixie Valley.

Sim.

5. Reactive transport simulations

161 Table 2 Chemical composition of injected fluid (#A) and steady state chemical composition obtained at the modeled column outlet for reactive transport simulations 1, 2, and 3 (#B, C, and D, respectively; concentrations in mg/l, except for Mg in ␮g/l). The values of parameters estimated by numerical optimization (Al and Mg concentrations, and dilution/concentration factor cfact) are also shown in the right-hand side columns, when computed.

L. Peiffer et al. / Geothermics 51 (2014) 154–169

L. Peiffer et al. / Geothermics 51 (2014) 154–169

gases from Goff et al. (2002). Gases were added back to the flashed water composition to simulate a pre-flashed composition. Then, this water was allowed to react (“batch” reaction simulation, within one model grid block, without flow) until it reached chemically steady conditions, near equilibrium with the minerals defined in the lopolith (deepest) model layer at a temperature of 250 ◦ C, representative of bottom-hole temperatures measured in this geologic unit. Finally, the resulting water composition (#A, Table 2) was injected at the bottom of the modeled column and allowed to react with wallrock minerals as it ascends and cools on its way up to the surface. For these simulations, boiling was suppressed by maintaining a pressure above vapor saturation. The resulting “synthetic” waters were “sampled” from the model top grid block and subjected to geothermometry calculations using both GeoT and classical geothermometers (Table 2).

quartz

A

5.3. Non-reequilibrated fluid (simulation #2) The extent of re-equilibration depends on the fluid injection rate and the reactive surface areas of the mineral phases (because these directly affect reaction rates, Eq. (1)) (e.g., Wanner et al., 2013, 2014). To minimize re-equilibration, the second simulation was run with a fast flowing fluid (25 m/day) having the same composition as in simulation #1. Moreover, the reactive surface areas in this simulation were set four orders of magnitude lower than the base-case values used in the previous simulation. This setup represents the fast upflow of a deep fluid along a fracture, in contrast with simulation #1. In this case the fluid is predicted to become supersaturated with respect to quartz and microcline along the column (Fig. 5A). This is also the case with albite, but only above ∼140 ◦ C. The fluid becomes supersaturated with respect to amorphous silica and chalcedony below ∼100 ◦ C and ∼210 ◦ C, respectively. Clinochlore remains close to chemical equilibrium within the gabbroic and basaltic layers, but then drops below saturation at shallower depths. The fluid is predicted to remain undersaturated with respect to montmorillonite, laumontite and clinoptinolite along the entire column. Consequently, the concentrations of Na, SiO2 and K do not vary significantly along the column (Fig. 5B) and their concentrations at 70 ◦ C reflect concentrations in the deep reservoir fluid. However, the modeled Al concentration at

chalcedony silica-am albite-lo

0 microcline

-1

clinochlore montmorillonite-ca

-2 laumontite

-3

clinoptinolite-Na

70

100

130

160

190

220

250

Temperature (°C)

B

1.E+03

1.E+02 1.E+01

5.2. Re-equilibrated fluid (simulation #1)

1.E+00

Concentration

As a first simulation, a slow injection rate was specified (<0.01 m/day) to promote the full re-equilibration of the ascending deep fluid with the wall-rock minerals. Albite, microcline and quartz remain near equilibrium with the fluid along the entire column, while clinochlore, laumontite and montmorillonite maintain equilibrium only in the geological layers where they are present. Na–clinoptinolite remains below saturation, as well as amorphous silica and chalcedony. The compositions of the ‘synthetic’ fluid sampled at depth (250 ◦ C) and at the surface (70 ◦ C) are shown in Table 2 (#A and #B, respectively). In this case, multicomponent geothermometry computations applied to compositions #A and #B reflect the sampling temperature (250 ◦ C and 70 ◦ C, respectively; Table 2). The Na–K geothermometer from Fournier (1979) somewhat underestimates the temperature (231 ◦ C and 54 ◦ C, Table 2), while the Na–K geothermometer from Giggenbach (1988) is more accurate (245 ◦ C and 75 ◦ C). The quartz geothermometer results are reasonable (238 ◦ C and 74 ◦ C), while Na–K–Ca and K–Mg give unreliable temperature estimates (272 ◦ C and 32 ◦ C, and 437 ◦ C and 229 ◦ C). This simulation represents the worst scenario for the use of solute geothermometers, because the fluid composition sampled at the surface has completely lost the deep reservoir chemical signature through re-equilibration on its way up to the surface.

2

1

log(Q/K)

162

1.E-01 SiO2 (mg/l)

1.E-02

Na (mg/l)

1.E-03

K (mg/l)

1.E-04

Al (mg/l) Mg (µg/l)

1.E-05 1.E-06 1.E-07 70

100

130

160

190

220

250

Temperature (°C) Fig. 5. (A) Mineral saturation indices computed during the reactive transport simulation of a fast-ascending fluid (simulation #2, Table 2). (B) Corresponding evolution of SiO2 , Na, K, Mg and Al concentrations during the simulation.

70 ◦ C is significantly lower than in the deep fluid (5.7 × 10−7 mg/l vs. 0.32 mg/l), and conversely the Mg concentration at 70 ◦ C is significantly higher than in the reservoir (1.1 × 10−3 vs 7 × 10−4 mg/l; Table 2, #C). The Al concentration decrease along the column due to the precipitation of microcline and albite (until ∼140 ◦ C), while Mg is mainly controlled by the dissolution of clinochlore and montmorillonite (Fig. 5). The evolution of the relative change in mineral abundances within the lopolith reservoir layer is shown in Fig. 6. Clinochlore dissolves (showing a negative volume fraction change) but only slightly, while quartz, albite and microcline are precipitating. The amount of dissolved clinochlore is much smaller than the precipitated amounts of the three other minerals (by 3 orders of magnitude) but still has a significant impact on the aqueous Mg concentration because of the low dissolved concentration of this element (Fig. 5). Similarly, the aqueous Al concentration is also quite sensitive to the precipitation of albite and microcline upon cooling. In contrast, the concentrations of SiO2 , Na and K are comparatively much higher and essentially not affected by the precipitation of these feldspars (Fig. 6). Thus, the dissolved concentration of trace elements like Al and Mg are quite sensitive to dissolution–precipitation processes. In contrast, the dissolved concentrations of major elements remain essentially unaffected when small quantities of these elements are removed by mineral precipitation or added by dissolution. For this reason, the classical Na–K and SiO2 geothermometers give reasonable temperature estimates (Na–KFournier1979 : 231 ◦ C, Na–KGiggenbach1988 : 245 ◦ C, SiO2 Fournier&Potter1982 : 237 ◦ C). However, because the Al and Mg concentrations at the surface differ significantly from reservoir conditions, the multicomponent geothermometry approach cannot predict reliable temperatures. The estimated temperature (68 ◦ C) using the simulated fluid composition at 70 ◦ C (including simulated Al) underestimates the reservoir temperature (Fig. 7A). However, using GeoT coupled with numerical optimization to estimate input Al and Mg concentrations

L. Peiffer et al. / Geothermics 51 (2014) 154–169 0,E+00

-5,E-08

A

-1,E-07 -2,E-07 clinochlore

Relative change in mineral abundance

-2,E-07

-3,E-07 -3,E-07 -4,E-07 -4,E-07 210

220

230

240

250

Temperature (°C) 4,E-04 4,E-04

B

3,E-04 quartz

3,E-04

albite-lo microcline

2,E-04 2,E-04 1,E-04 5,E-05 0,E+00 210

220

230

240

5.4. Mixing with superficial saline waters (simulation #3)

Fig. 6. Modeled relative change in mineral volume abundances (dimensionless) inside the lopolith layer (simulation #2, Table 2).

(0.40 and 0.06 ␮g/l, respectively), a reliable reservoir temperature is estimated (250 ◦ C, stdev 4 ◦ C; Fig. 7B). The optimized case shows much better clustering of the log(Q/K) curves at the computed temperature than the non-optimized case. As would be expected, the approach works best when minerals known to be at or near equilibrium in the reservoir are selected. Otherwise, the curves of saturation indices show a large spread 4

Simulation #2

4

B

Simulation #2 (opt. Al-Mg) 3

2

2

2

1

1

1

0 -1

0

-1

microcline

albite-lo

clinochlore

-3

2.0 1.8 1.6 1.4 1.2 68°C 1.0 0.8 RMED 0.6 RMSE 0.4 Computed T SDEV 0.2 MEAN 0.0 60 80 100 120 140 160 180 200 220 240 260 280 300 Temperature (°C)

Log(Q/K) Statistics

-4

quartz

microcline

albite-lo

clinochlore

-3

-4

-4

2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0

RMSE SDEV MEAN

Computed T

250°C

100 120 140 160 180 200 220 240 260 280 300 Temperature (°C)

Simulation #2 (opt. Al-Mg, full min.list)

0

2.0 1.8 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0

RMED

C

-1 -2

-2 quartz

Log(Q/K) Statistics

-3

Log(Q/K)

3

-2

Log(Q/K) Statistics

The mixing of deep fluids with superficial saline water can seriously affect results from classical geothermometers. As an example, a 0.15 m NaCl solution is injected into the column at an arbitrary temperature of 120 ◦ C. This solution has the same salinity as the Dixie Valley marsh waters (Goff et al., 2002). The injection rate is half the fracture flow rate specified for simulation #2, thus mixing with the original injected fluid and diluting all species’ concentration other than Na and Cl by a factor of 1.5. Consequently, the Na and Cl content of the rising deep fluid is predicted to increase while the concentrations of K and SiO2 decrease proportionally to

3

Log(Q/K)

Log(Q/K)

A

in temperature (Fig. 7C), and the optimization process may yield several minima and an erroneous temperature, although in this particular case the computed temperature (232 ◦ C) is close to the reservoir temperature. Note that the use of minerals with both prograde and retrograde solubilities greatly constrains the optimization procedure. In the above example, the estimated Al concentration by numerical optimization is similar to the Al concentration of the synthetic water at 250 ◦ C (Table 2, #A), and falls between the measured “dissolved” (ionized, filtered <0.2 ␮m) and “total” (filtered <0.45 ␮m) Al concentrations reported by Goff et al. (2002) for Dixie Valley geothermal well waters (e.g. for well 73-7, dissolved Al 0.063 mg/l and total Al 1.04 mg/l; Table 3). This observation reflects the loss of Al, from the upwelling fluid, by the precipitation of aluminosilicates (in the model) and/or colloidal phases (in natural systems). These phases, or at least a fraction of them, might be transported together with the fluid to the surface, in which case “total” Al concentrations probably would reflect better the Al concentration in the deep reservoir. In contrast, the Al concentration in the re-equilibrated fluid (synthetic water #B, Table 2) is similar to observed “dissolved” Al concentrations. This suggests that the analyzed “dissolved” Al concentrations in fluid samples from Dixie Valley geothermal wells reflect some degree of Al precipitation by cooling, which is further supported by analyses of scale in these wells revealing the presence of clays in addition to silica phases (Bruton et al., 1997).

250

Temperature (°C)

4

163

calcite microcline pyrite clinoptinolite-Na montmorillonite-Ca

quartz albite-lo clinochlore laumontite goethite

RMED RMSE SDEV

232°C

MEAN

Computed T

100 120 140 160 180 200 220 240 260 280 300 Temperature (°C)

Fig. 7. Mineral saturation indices (log(Q/K)) computed for the synthetic water from simulation #2. (A) Non-optimized case and (B) the concentrations of Al and Mg are estimated by numerical optimization. (C) Similar to B but considering all minerals present in the entire modeled column. Plots of respective median (RMED), root mean squared (RMSE), standard deviation (SDEV) and average (MEAN) of absolute log(Q/K) values are also shown. The estimated temperature is given by the minimum RMED.

164

L. Peiffer et al. / Geothermics 51 (2014) 154–169

Table 3 Chemical composition of selected representative thermal waters from Dixie Valley (Goff et al.,2002). The sampling temperature T (◦ C) and the steam fraction (sf) data are also reported when available. Samples Wells 73-7 76-7 74-7 82A-7 63-7 73B-7 27-33 37-33 28-33 45-33 pre 73-7 pre 74-7 pre 76-7 pre 32-18 pre 65-18 pre 45-14 62-21 Springs Hyder Fault Line Lower Ranch Jersey McCoy Sou Dixie Big Horn

T (◦ C)

sf

pH

SiO2

Na

K

∼165 163 n.m. n.m. n.m. 174 ∼165 n.m. n.m. n.m. n.m. n.m. n.m. n.m. n.m. 125 76

0.16 0.18 0.16 0.16 0.15 0.15 0.16 0.16 0.16 0.17 0.20 0.20 0.19 n.m. n.m. n.m. n.m.

9.0 9.1 9.1 9.1 9.0 9.0 8.8 9.2 9.4 9.1 9.0 9.1 9.2 7.6 8.9 7.0 6.9

580 599 586 556 516 511 627 621 531 589 548 574 563 484 417 285 172

508 474 500 495 510 500 423 431 412 370 380 413 403 406 440 432 513

74 70 72 73 77 74 67 69 66 59 59 62 54 43 41 41 17

75 29 41 59 46 73 82 21

n.m. n.m. n.m. n.m. n.m. n.m. n.m. n.m.

6.4 7.0 7.0 7.0 7.0 7.0 7.2 7.9

58 42 41 109 36 63 105 36

357 162 141 261 185 162 194 427

21 12 11 26 9.1 28 4.9 4.5

Ca

Mg

9.0 8.5 9.2 10 8.7 8.4 7.7 7.2 7.2 1.3 1.2 1.1 1.5 2.1 1.2 23 6.1 43 67 38 24 79 112 11 87

Al(t)

Al(d)

Fe

Cl

SO4

HCO3

0.02 0.03 <0.01 <0.01 <0.01 <0.01 <0.01 0.02 0.03 0.04 <0.01 <0.01 <0.01 <0.01 <0.01 0.04 0.41

1.04 1.12 1.13 1.02 1.03 0.97 1.44 0.99 1.36 1.52 0.99 1.10 1.19 0.75 0.69 0.23 0.07

0.06 0.05 0.08 0.04 0.02 0.04 0.04 0.02 0.04 n.a. n.a. n.a. n.a. n.a. n.a. n.a. 0.09

0.04 <0.01 0.01 0.02 <0.01 <0.01 <0.01 0.02 <0.01 n.a. n.a. n.a. n.a. n.a. n.a. 0.07 0.24

594 524 584 575 560 561 443 475 441 320 363 396 402 428 404 481 80

207 201 204 212 214 216 183 191 199 149 150 159 158 150 162 195 219

161 173 183 164 153 153 188 172 178 311 291 309 286 223 334 101 836

10 19 13 2.8 30 21 0.12 47

<0.02 <0.01 <0.01 0.09 <0.01 <0.01 0.03 0.03

n.a. n.a. n.a. n.a. n.a. n.a. n.a. n.a.

0.07 <0.01 <0.01 0.05 <0.01 0.27 <0.01 0.03

45 32 30 49 228 77 161 707

118 133 68 147 199 379 139 181

735 450 369 426 265 258 77 181

Al(t) is total 0.45 ␮m filtered Al concentration; Al(d) is dissolved 0.2 ␮m filtered Al concentration. All concentrations in mg/l except for Mg in ␮g/l.

best fit with a set of thermal water compositions (Giggenbach, 1988). However, because of the high sensitivity of Mg concentrations to re-equilibration, it is probable that the Mg concentrations of a thermal water sampled at the surface do not reflect deep reservoir concentrations, because small amounts of precipitation/dissolution of Mg-bearing minerals can significantly affect

4

Simulation #3 (opt. Al-Mg-cfact) 3

Log(Q/K)

2 1 0 -1 -2 -3

Log(Q/K) Statistics

the dilution factor (Table 2, #D). In this case, the Na–K and quartz geothermometers applied to the modeled water at 70 ◦ C significantly underestimate the reservoir temperature (Na–KFournier1979 : 129 ◦ C Na–KGiggenbach1988 : 149 ◦ C; SiO2Fournier&Potter1982 : 203 ◦ C). However, using GeoT coupled with numerical optimization to estimate not only the Al and Mg concentrations but also a dilution factor, good results are obtained. In this case the computed reservoir temperature estimate is 250 ◦ C, and the estimated Al and Mg concentrations (0.12 and 0.07 ␮g/l, respectively) and dilution factor (1.6) have reasonable values. The standard deviation associated with the GeoT temperature estimate is higher than for simulation #2 (11 ◦ C vs. 4 ◦ C) because the albite saturation index crosses the zero value at a higher temperature than in the previous case (Fig. 8) because the Na concentration is affected by the addition of NaCl. For all three simulations, the classical Na–K–Ca geothermometer yields unreliable temperature estimates (e.g. for simulation #2, 217 ◦ C; Table 2). This is because Ca concentrations in the model are affected by equilibration with calcite along the entire model column, and therefore are not representative of the deep reservoir. The K/Mg geothermometer (Giggenbach, 1988) significantly overestimates temperatures (e.g., for simulation #2, 308 ◦ C). This a consequence of the low Mg concentrations obtained in the three simulations, being lower than the detection limit (<0.01 mg/l; Table 2). The K/Mg correlation of Giggenbach (1988) was derived from a two dimensional projection (log([K]2 /[Mg]) vs. temperature) of the stability field of the minerals muscovite, clinochlore and K-feldspar. As mentioned by this author, this kind of diagram only indicates the potential stability field of those minerals and not necessarily the formation of these minerals (e.g., the SiO2 or Al concentrations may be too low for these minerals to form). The multicomponent geothermometry approach solves this problem because it takes into account the concentrations of all dissolved elements necessary to assess whether certain minerals belong or not to the equilibrium assemblage. Furthermore, the Giggenbach K/Mg geothermometer was derived from a parallel line 0.3 log units above the coexistence line of K-feldspar, clinochlore and muscovite. This regression line was chosen because it offered the

quartz

microcline

albite-lo

clinochlore

-4 2.0 1.8 1.6 1.4 RMED 1.2 RMSE 1.0 SDEV 0.8 MEAN 0.6 252°C 0.4 Computed T 0.2 0.0 100 120 140 160 180 200 220 240 260 280 300

Temperature (°C) Fig. 8. Mineral saturation indices (log(Q/K)) computed for the synthetic water from simulation #3. The concentration of Al and Mg, and the dilution/concentration factor (cfact) are estimated by numerical optimization. Respective stastitical indices for log(Q/K) values are also shown (see caption of Fig. 7 for abbreviations).

L. Peiffer et al. / Geothermics 51 (2014) 154–169

the dissolved Mg content (e.g., as in simulation #2). This may explain the discrepancy between K/Mg temperature estimates and modeled temperatures in our case. Giggenbach (1988) also correctly points out that discrepancies due to uncertainties in mineral thermodynamic data cannot be ruled out. 6. Field data analysis and discussion The GeoT code coupled with numerical optimization was applied to the different groups of Dixie Valley waters identified earlier (Fig. 3). The parameters estimated by optimization were the Al and Mg concentrations; the dilution/concentration factor (‘cfact’, representing concentration when its value is <1); and also the steam fraction (‘sf’, the fraction of gas in the total discharge) when not reported by Goff et al. (2002). For waters without gas analyses (e.g., springs), the following average gas composition was estimated from analyses of gas samples from geothermal wells (Goff et al., 2002) to reconstitute the ‘pre-boiled’ fluid composition: 99.8 mol% H2 O (wet gas); and 95.2 mol% CO2 , 1.081 mol% H2 S, 0.826 mol% CH4 , and 0.0827 H2 (dry gas). The chemical compositions of the selected waters subjected to geothermometry calculations, as well as sampling temperatures, are shown in Table 3. Corresponding multicomponent (GeoT–iTOUGH2 estimates) and classical geothermometry results, together with parameters estimated by optimization, are listed in Table 4. The standard deviations (Tstdev ) associated with GeoT estimates are also presented in Table 4. These values express a measure of the range of equilibration temperatures obtained with each individual mineral. It should be noted that temperature estimates from any solute geothermometry method should best be considered to have an uncertainty in the 10 ◦ C range. When several waters could be identified as belonging to a same group, GeoT–iTOUGH2 was run using the analyses of these waters simultaneously to better constrain the optimization process. The average computed temperatures for these groups of waters are also reported in Table 4 and discussed below. For every GeoT simulation coupled with numerical optimization (using single or multiple water analyses of wells or springs), the best results were obtained by limiting the number of minerals to the set of four key minerals present in the lopolith reservoir: quartz, albite, microcline, and clinocholore (chlorite). These minerals are also present in the other geological formations underlying Dixie Valley, with the exception of chlorite, which is usually not present in the alluvial-colluvial deposits (Lutz, XRD unpublished data). Several tests were performed using montmorillonite instead of clinochlore, but the saturation indices curve of this mineral rarely converged with other minerals, because of either a lack of equilibrium with the water samples, or uncertain thermodynamic data (which is expected for minerals of variable composition such as clays). Simultaneous optimization of multiple waters proved to be difficult when including a larger number of minerals, because it significantly increased the number of local (false) minima of the objective function. For this reason, further investigations are being conducted to evaluate efficient alternative minimization procedures that could be used successfully with a large number of minerals when multiple waters are processed simultaneously. 6.1. Production wells Production wells from Section 7 were first considered. Numerical optimization of Al and Mg concentrations and the dilution/concentration factor were performed using analyses of water and gas samples from 6 wells from Section 7 (#1, Table 4). Steam fractions (0.15–0.18) reported in Goff et al. (2002) were used as

165

input into GeoT to reconstruct the deep fluid composition. Doing so resulted in an estimated average temperature of 250 ◦ C, close to the range of measured downhole temperatures in Section-7 wells (240–250 ◦ C). The dilution/concentration factor was optimized to take into account the reservoir salinity increase caused by reinjection. The optimized value suggests an increase in reservoir salinity of ∼15–25% after production, consistent with observations by Kennedy et al. (1999). In general, the Al concentrations estimated by optimization (Table 4) are fairly consistent with the reactive transport model results presented earlier (0.32 mg/l, simulation 1, #1), and within the range of measured “dissolved” (<0.2 ␮m filtered) and “total” (0.45 ␮m filtered) Al concentrations (Table 3). However, the optimized Mg concentrations are much lower than field analyses (Table 4) but significantly higher than previously modeled values (∼0.07 ␮g/l). It should be noted that Mg concentrations are constrained only by one Mg mineral in this case, clinochlore. The stability of this mineral is more sensitive to pH compared to albite or microcline. Therefore, slight differences in the pH of production well waters and modeled synthetic waters cause noticeable differences in Mg concentrations. Waters from three Section-33 wells were processed in a similar manner. The average GeoT estimated temperature (247 ◦ C) and values of optimized parameters (Table 4, #2) are similar to results obtained with data from Section-7 wells and reported measured downhole temperatures (Table 4, #1). The classical quartz and Na–K geothermometers estimates for both Section 33 and 7 wells (average respectively of 250–257 ◦ C and 253–260 ◦ C) are in reasonably good agreement with GeoT results. However, the Na–K–Ca and K–Mg geothermometers give somewhat lower temperatures for both sections (231–236 ◦ C). Reinjected fluids are a mixture of separated brines and superficial waters with higher content in Ca and Mg, possibly causing the last two geothermometers to yield lower temperatures. 6.2. Pre-production wells Pre-production wells from Sections 33 and 7 also show similar GeoT estimated temperatures (256 ◦ C and 247 ◦ C). The classical geothermometer estimates in this case are mostly within the standard deviation associated with the GeoT estimates (247–261 ◦ C), except for the K/Mg geothermometer (#3 and #4, Table 4). Waters from Section 18 pre-production wells were also examined, considering the only two available analyses reported by Goff et al. (2002) for these wells (Table 4, #5). The steam fraction (of the total steam + water discharge) was not reported for these wells, therefore this parameter was estimated by optimization, together with Al and Mg concentrations. No dilution or concentration was considered (‘cfact’ set to 1) because pre-production well waters (corrected for steam loss) should represent the original reservoir brine. The reservoir temperature estimated by GeoT in this case (246 ◦ C) is similar to that estimated for Sections 7 and 33 production wells. The optimized steam fraction for these wells is within the range of steam fraction values observed for other Dixie Valley wells (Table 3). The quartz, Na–K and Na–K–Ca geothermometers in this case give somewhat lower temperatures (Table 4) that are more in line with measured downhole temperatures for Section 18 wells reported by Waibel (1987) (225 ◦ C). The divergence between the classical geothermometers, measured downhole temperatures and GeoT estimates for the Section-18 wells could be the result of a hotter fluid coming from the underlying lopolith reservoir (the bottom of well 3218 is located above this reservoir, in the basaltic rocks, Tmb). GeoT-estimated temperatures may represent the equilibrium temperature of fluids originating from the lopolith. Examining the water isotopic composition of the pre-production wells from Sections 33, 7 and 8, there is almost no variation in pre-flash ␦D and

166

L. Peiffer et al. / Geothermics 51 (2014) 154–169

Table 4 Results of solute geothermometry using classical geothermometers, optimized multicomponent geothermometry with GeoT, and values of parameters estimated by numerical optimization. The classical geothermometers are applied to the reconstructed deep fluids using GeoT. Bottomhole temperature (BHT) are also reported when available in Goff et al. (2002). Sample names ending by -7, -33, -18 stand for well from Sections 7, 33, 18; “pre” designates pre-production wells. #

n

Wells 1

6

2

Average 3

3 4

Average 1 3

5

Average 2

6 7

Average 1 1

Springs 8

9 10 11 12

4

Average 1 1 1 1

Samples

BHT

73-3 76-7 74-7 82A-7 63-7 73B-7

n.m. n.m. 244 241 241 n.m.

27-33 37-33 28-33

243 n.m. 246

45-33 pre 73-7 pre 74-7 pre 76-7 pre

n.m. n.m. n.m. n.m.

32-18 pre 65-18 pre

225 225

45-14 62-21

196 n.m.

Hyder Fault Line Lower Ranch Jersey

– – –

McCoy Sou Dixie Big Horn

– – – –

Tquartz a

TNa-K

TNa-K-Ca

255 255 256 250 243 243 250 264 263 245 257 255 244 249 248 247 235 224 229 200 164

252 252 251 252 255 253 253 260 261 260 260 261 258 254 244 252 223 210 217 213 138

232 231 230 231 235 233 232 235 237 235 236 254 252 252 241 248 224 221 223 194 153

108 93 92 141 108 87 112 139 87

174 192 199 218 196 163 270 122 79

158 154 161 187 165 136 195 123 87

TK–Mg

241 230 – – – – 236 – 237 224 231 212 – – –

TgeoT

Tstdev

Optimized parameters Al

Mg

cfact

sf

0.85 0.80 0.85 0.81 0.87 0.86 0.84 0.85 0.85 0.85 0.85 – – – –

– – – – – –

– –

198 124

252 252 252 252 248 244 250 248 248 244 247 256 248 244 248 247 248 244 246 224 176

6 6 6 26 7 3 5 7 7 3 5 8 11 13 12 12 8 3 6 2 3

0.42 0.56 0.42 0.62 0.35 0.42 0.47 0.40 0.35 0.44 0.39 0.35 0.29 0.25 0.34 0.29 0.32 0.52 0.42 0.43 0.19

0.58 0.49 0.49 0.49 0.84 1.06 0.66 0.47 0.46 0.51 0.48 0.90 0.69 0.70 0.49 0.62 0.94 0.42 0.68 6.21 7.90

– –

0.18 0.16 0.17 0.08 0.1

84 64 67 107 81 53 83 106 35

180 180 180 188 182 180 184 200 200

2 5 4 7 5 2 8 10 11

0.03 0.02 0.05 0.06 0.04 0.03 0.08 0.09 0.03

0.75 0.98 0.89 0.53 0.78 7.57 2.20 1.96 1.35

3.08 4.59 3.94 1.92 3.38 4.73 2.66 2.39 5.80

0.02 0.02 0.03 0.02 0.02 0.01 0.02 0.02 0.02

– –

– – – – – – –

Quartz, Na–K equations are respectively from Fournier and Potter (1982; conductive cooling formula) and from Fournier (1979); Na–K–Ca from Fournier and Truesdell (1973) and K–Mg from Giggenbach (1988). The Al concentration is in mg/l, and the Mg concentration in ␮g/l. Temperature is in degrees Celsius. sf, steam fraction; n, numbers of samples processed simultaneously in the optimization. a Quartz geothermometer was calculated taking into account the measured or optimized steam fractions. When Mg concentrations are below limit detection, the K–Mg geothermometer is not calculated.

␦18 O values (Fig. 3C and D), suggesting that those fluids could have a common origin.

6.3. Other geothermal and on-site wells (‘ogo’) wells 45-14 and 62-21 Because of their difference in chemical compositions compared to production wells (Fig. 3), the water analyses from ‘ogo’ wells 45-14 and 62-21 were processed separately. The GeoT estimate for well 45-14 (224 ◦ C) is higher than classical geothermometer estimates by up to ∼30 ◦ C, the latter being more in line with measured maximum downhole temperatures about 196 ◦ C (Table 4, #6). Steam fractions were not reported for this well. The steam fraction estimated by optimization (0.08) is lower than at other production wells, but similar to steam fractions reported for other non-productive wells (e.g., well 27-32: 0.05; Goff et al., 2002). The optimized Al concentration (0.43 mg/l) is similar to measured “total” (0.45 ␮m filtered) concentrations (∼0.23 mg/l) and previous modeling results (Table 2). However, the Mg concentration is an order of magnitude higher (6.21 ␮g/l) than concentrations optimized previously for other wells (Table 4). The data from well 45-15 show discrepancies between GeoT results, classical geothermometer estimates, and measured temperature that are similar in nature to the case of Section 18 pre-production wells. As for Section 18 wells, this could reflect higher temperature fluids at depths below the well bottom.

For well 62-21, GeoT yields a temperature estimate of 176 ◦ C (Table 4, #7). The steam fraction (0.1) also had to be estimated by optimization for this well. The fluid entry in this well was reported to occur at the contact between the gabbro and underlying Triassic metasediments. From isotherms in Fig. 2, a temperature near 170 ◦ C prevails at the contact between both lithologies in this well, which would support the GeoT results. However, classical geothermometers yield temperatures that are lower by up to ∼50 ◦ C. The Al and Mg concentrations estimated by optimization are similar to results for well 45-14 (Table 4). 6.4. Hot springs Steam fraction is not known for springs, therefore this parameter was estimated by optimization. Furthermore, springs may have been diluted, so the dilution factor ‘cfact’ was also estimated by optimization, together with Al and Mg concentrations (Table 4). We first investigated the northeastern thermal springs (Hyder–Jersey–Lower Ranch–Fault Line) that are characterized by a low Cl–Na ratio (Fig. 3), considering the same four minerals as previously (quartz, albite, microcline, and clinocholore). The optimization was realized considering the four spring analyses (Table 4, #8). The average computed temperature for these springs (182 ◦ C) differs significantly from classical geothermometers estimates. This temperature is more than ∼70–100 ◦ C higher than that given by the quartz and K–Mg geothermometers, and within the range of temperatures (∼±30 ◦ C) obtained with the Na–K and Na–K–Ca

L. Peiffer et al. / Geothermics 51 (2014) 154–169

geothermometers (using the reconstructed water composition) (Table 4, #8). The difference between the classical geothermometers and GeoT results most likely reflects dilution and mixing with shallow dilute waters richer in Ca and Mg. Note that by multiplying the measured SiO2 concentrations by the optimized dilution factor (‘cfact’) values, the quartz geothermometer estimates become closer to the Na–K geothermometer results. McCoy springs belong to the intermediate Cl–Na trend (Fig. 3A), and yields a GeoT estimated temperature of 180 ◦ C that is similar to other north-eastern springs. Classical geothermometers give lower and divergent estimates ranging from 53 ◦ C to 163 ◦ C (Table 4, #9). Sou springs are characterized by distinct Cl/Na ratios (that fall between the low and intermediate Cl/Na trends) but isotopic compositions similar to other north-east springs. These waters yield a similar GeoT temperature (184 ◦ C), but diverging classical geothemometers estimates (83–270 ◦ C; Table 4, #10). The waters of Dixie and Big Horn springs follow, respectively, an intermediate and halite Cl–Na trend (Fig. 3A). Because of their distinct salinities and isotopic compositions, as well as different sampling temperatures (Table 3), waters from these springs were processed individually. Nevertheless, they are both characterized by a similar GeoT temperature (200 ◦ C) that is much higher than the classical geothermometers estimates (35–140 ◦ C; Table 4, #11–12). These springs are located close to the Humboldt Salt Marsh. Therefore, mixing between the deep upwelling geothermal fluid and superficial saline brines from the marsh area, or salt leaching, could explain the departure between GeoT estimates and the classical geothermometers. The dilution factors estimated by optimization for the spring waters are elevated (from about 2 to 6, Table 4), suggesting that these waters have undergone a significant dilution on their way to ground surface. These dilution factors are consistent with the spring compositions: the higher optimized dilution factor corresponds to the more dilute spring composition. The steam fractions estimated by optimization for the springs are between 0.01 and 0.03, suggesting conductive cooling rather than boiling. This is confirmed by the ␦18 O–␦D values of the springs that are typically on the DMWL (except for Hyder springs) and do not reveal any shift due to boiling fractionation (Fig. 3B). The optimized Al concentrations (0.02–0.09 mg/l) are close to the “total” Al measured concentration (<0.01–0.09 mg/l). The optimized Mg concentrations (0.0.53–7.57 ␮g/l) are lower than the measured ones (0.12–47.4 mg/l) and similar to the values modeled previously. 6.5. Two major reservoirs under Dixie Valley? The available water chemical and isotopic data from Dixie Valley (Nimz et al., 1999; Goff et al., 2002) clearly points out the complexity and variety of water chemical composition encountered at this geothermal area. Our review of these data together with our new geothermometry results support to the existence of at least two main reservoirs in this area. Geothermal well waters sampled from Sections 33, 7, and 18 show similar chemical and isotopic compositions and are considered to belong to a common reservoir located along the Stillwater range normal fault system, in the Jurassic Humboldt Lopolith. This reservoir could be compartmentalized in distinct smaller reservoirs with slightly different compositions as proposed by Kennedy et al. (1999). Equilibrium temperatures computed using pre-production and production well water compositions were estimated to be in the 240–260 ◦ C range. The increasing salinity trend from the north to the south observed in pre-production well waters is difficult to explain. Waters in this reservoir could have originated from a Late Pleistocene lake that infiltrated through the Stillwater fault system (Nimz et al., 1999). It is also possible to explain the salinity increase

167

of spring waters by mixing of a deep fluid with superficial saline brines and/or by leaching of salt from the marsh area. However, for well waters, such mixing or leaching would be unlikely. Instead, the salinity increase could reflect an inherited characteristic of the time when the geothermal reservoir was formed, when the lake brines infiltrated along the Stillwater fault to the south, close to the actual marsh. Other sampled waters from well 45-14, Dixie and Big Horn springs (south of the production area) show compositions with intermediate to high Cl/Na ratios, similar to the production well waters, although their trend in isotopic composition is difficult to interpret and cannot be used to clearly conclude that these waters originate from a similar reservoir. Estimated temperatures for these waters range from 200 to 225 ◦ C. Northeastern Hyder, Lower Ranch, Fault Line, and Jersey springs, together with well 62-21, constitute another water group with distinct chemical and isotopic compositions compared to the other geothermal wells. This group may be fed by a second reservoir with temperatures estimated at 175–190 ◦ C. This temperature range corresponds to the measured isotherms at the contact between the Jurassic lopolith and the Triassic marine carbonates (Fig. 2). The elevated total dissolved carbonate content in these waters could originate from the Triassic carbonates, and the lack of Cl-enrichment (low Cl/Na ratio) compared to waters further southwest might be related to absence of the Dixie Pluvial Lake in this part of the valley (Thompson and Burke, 1973). Well 62-21 intercepts a normal fault on the eastern side of the valley (Fig. 2). This fault is part of a fault system that extends both to the southwest and northeast in the valley, approximately parallel to the Stillwater Range (Fig. 1B). Hyder, Lower Ranch, Fault Line and Jersey springs at the northeast portion of Dixie Valley are likely surface expressions of this fault system. It is known by high-resolution aeromagnetic survey that Hyder springs occur at the intersection of two faults: a deep (∼1 km depth) NW trending fault intersecting a shallow (<100 m depth) fault at the position of the springs (Blackwell et al., 2007a). By their similarity in estimated GeoT temperatures, chemical composition (except in Cl) and geographic position, McCoy springs probably belong to the same reservoir. Their enrichment in Cl is attributed to the mixing with saline superficial brines or the leaching of salts. Sou springs are more enriched in Ca and SO4 (Table 3), and are characterized by a slightly higher Cl/Na compared to other northeastern springs. Although they show similar GeoT temperatures, it is unclear whether or not they are related to this reservoir. Exploration holes were drilled in the 1970s in the vicinity (∼1 km) of Hyder and Sou springs to a depth of 900 m beneath the surface. From a depth of 450 m, recorded temperatures were ∼68–70 ◦ C, increasing only a few degrees at the bottom of the borehole. Such temperature gradients were not high enough to pursue commercial well drilling (Dick Benoit, personal communication). The lack of significant temperature gradient observed close to the springs does not rule out the hypothesis of a hot reservoir at depth. The springs might be localized systems with very restrained thermal anomalies that are connected by faults to a unique reservoir at depth. Furthermore, He isotopic compositions of Hyder, Jersey, McCoy and Sou springs are similar and describe a trend representing the mixing between a young groundwater endmember and a deep enriched fluid (Kennedy and van Soest, 2006). Therefore, the hypothesis of one deep common reservoir for these springs is reasonable, although it cannot be proven at this time.

7. Summary and conclusions Solute geothermometry was investigated at Dixie Valley using classical geothermometers and a new approach coupling multicomponent geothermometry with numerical optimization

168

L. Peiffer et al. / Geothermics 51 (2014) 154–169

following Spycher et al. (2014). This approach builds on earlier studies by Reed and Spycher (1984) and Pang and Reed (1998), and allows the estimation of unknown or poorly constrained parameters that affect temperature predictions, such as the concentration of key elements (e.g., Al and Mg), dilution amounts, and/or the amount of gas lost from geothermal fluids before these can be sampled for analysis. Insights on solute geothermometry were gained from results of reactive transport modeling simulations using TOUGHREACT V2.0 (Xu et al., 2011), showing that the concentrations of Al and Mg are significantly affected by precipitation–dissolution processes (even with minute amounts of reaction), whereas major elements are much less affected. The composition of synthetic waters from the reactive transport simulations were used as inputs into optimized multicomponent geothermometry computations. As would be expected, results were optimal when the minerals known to be at (or close to) equilibrium in the reservoir were selected for these computations. In general, temperatures computed for Dixie Valley geothermal wells correspond to the measured downhole temperatures, although in some cases (e.g., Section 18 pre-production wells) computed temperatures are higher, suggesting that hotter conditions could exist at greater depths. Steam fractions were estimated by the optimization procedure when unavailable, and the increased salinity of the geothermal reservoir due to evaporated brine reinjection was reasonably well predicted. For thermal springs, the dilution factors estimated by numerical optimization were significant, contrasting with geothermal well waters showing increased salinities caused by re-injection of evaporated brines. Al concentrations estimated by optimization for both well and spring waters are generally closer to the measured “total” (filtered <0.45 ␮m) concentrations than the “dissolved” (filtered <0.2 ␮m) concentrations. This suggests the transport of fine (<0.45 ␮m) Al precipitates or colloids with the uprising fluid to the surface, or the precipitation of Al from the solution during sampling. Therefore, at this geothermal site, analyzing the “total” (filtered <0.45 ␮m) Al concentration in the sampled fluid appears to more closely reflect the Al concentration in the reservoir than the “dissolved” (<0.2 ␮m) fraction. For Mg concentrations, the analytical issues differ because the mineral controlling solubility in this case is chlorite (clinochlore), which displays a retrograde solubility behavior. Therefore, as fluids become undersaturated with respect to this mineral upon cooling, driving dissolution from the wall rock, the measured concentrations of Mg are likely to be overestimated, and thus are not likely to reflect reservoir conditions. Our investigation also suggests that classical geothermometers may yield underestimated temperatures with spring waters, because of mixing with saline Na–Cl brines from the marsh area and shallow dilute Ca–Mg rich waters. Our review of available hydrochemical data, together with geothermometry analyses, suggests that there could be two main geothermal reservoirs at the Dixie Valley geothermal area. The first reservoir has an estimated temperature of 240–260 ◦ C. It covers the area corresponding to the geothermal field but could extend towards the south-west where deep temperatures of 200–225 ◦ C were estimated. The second reservoir has an estimated temperature of 175–190 ◦ C and extends from well 62-21 to northeastern Hyder, Lower Ranch, Fault Line, and Jersey springs. The extension of these reservoirs, approximated by the distance between the wells and springs discussed throughout the study, is potentially much larger than the currently exploited geothermal area (Fig. 1A). Therefore, the geothermal energy potential of Dixie Valley could be much higher than at the energy currently produced from this geothermal area. The results presented in this study are more consistent with measured downhole temperatures than previous attempts to use multicomponent geothermometry with Dixie Valley geothermal

well waters (Bruton et al., 1997; Shevenell and De Rocher, 2005). This shows that combining multicomponent geothermometry with numerical optimization, using data from multiple locations when appropriate, can considerably increases the power and range of applicability of this method. Also, our study further points out the importance of selecting a mineral assemblage that is as representative as possible, and using adequate Al and Mg input concentrations, for successful application of this approach. Multicomponent geothermometry is not intended to replace classical geothermometers, but rather to supplement these geothermometers, and by doing so to increase confidence in temperature estimations. However, such approach cannot be applied carelessly and without a sound conceptual understanding of the area being studied. Also, as pointed out previously (Spycher et al., 2014), numerical optimization, by itself, can present a challenge, often requiring several trials using different methods and/or initial parameter values. For this reason, there remains room for improvement of the method by developing/testing alternative optimization procedures that are efficient and successful for systems considering a large number of waters and reservoir minerals. Acknowledgements This work was supported by the U.S. Department of Energy, Geothermal Technologies Program, Energy Efficiency and Renewable Energy Office, Award no. DE-EE0002765. We thank Susan Lutz for providing XRD analyses, and Dick Benoit and Lisa Shevenell for personal communications regarding Dixie Valley. We are also grateful to Patrick Dobson for a constructive review of the original manuscript. Reviews by S. Simmons and an anonymous reviewer are also greatly appreciated. References Barnes, R.B., 1975. The determination of specific forms of aluminum in natural water. Chem. Geol. 15, 177–191. Benoit, W.R., 1989. Carbonate scaling characteristics in Dixie Valley, Nevada geothermal wellbores. Geothermics 18, 41–48. Bergfeld, D., Goff, F.E., Janik, C.J., 2001. Elevated carbon dioxide flux at the Dixie Valley geothermal field, Nevada: relations between surface phenomena and the geothermal reservoir. Chem. Geol. 177, 43–66. Blackwell, D.D., Smith, R.P., Bergman, S., Goff, F., Kennedy, M., McKenna, J., Richards, M., Waibel, A., Wannamaker, P., 2007a. Description, synthesis, and interpretation of the thermal regime, geology, geochemistry and geophysics of the Dixie Valley, Nevada geothermal system. Nevada Bureau of Mining Geology, 250 pp. (internal report). Blackwell, D.D., Smith, R.P., Richards, M.C., 2007b. Exploration and development at Dixie Valley, Nevada: summary of DOE studies. In: Proceedings 22nd Workshop on Geothermal Reservoir Engineering, Stanford Univ. Report SGP-TR-183. Bruton, C., Counce, D., Bergfeld, D., Goff, F., Johnson, S., Moore, J., Nimz, G., 1997. Preliminary investigation of scale formation and fluid chemistry at the Dixie Valley geothermal field, Nevada. Geoth. Resour. Council Trans. 21, 157–164. Doherty, J., 2008. PEST – Model-independent Parameter Estimation. Watermark Numerical Computing, Corinda 4075, Brisbane, Australia http://www.sspa.com/pest/ Dobson, P.F., Kneafsey, T.J., Sonnenthal, E.L., Spycher, N., Apps, J.A., 2003. Experimental and numerical simulation of dissolution and precipitation: implications for fracture sealing at Yucca Mountain, Nevada. J. Contamin. Hydrol. 62–63, 459–476. Finsterle, S., Zhang, Y., 2011. Solving iTOUGH2 simulation and optimization problems using the PEST protocol. Environ. Model. Softw. 26, 959–968. Fournier, R.O., Truesdell, A.H., 1973. An empirical Na–K–Ca geothermometer for natural waters. Geochim. Cosmochim. Acta 37, 1255–1275. Fournier, R.O., 1979. A revised equation for the Na/K geothermometer. Geoth. Resour. Council Trans. 3, 221–224. Fournier, R.O., Potter, R.W., 1982. A revised and expanded silica (quartz) geothermometer. Geotherm. Resourc. Council Bull. 11, 3–12. Giggenbach, W.F., Stewart, M.K., 1982. Processes controlling the isotopic composition of steam and water discharges from steam vents and steam-heated pools in geothermal area. Geothermics 11, 71–80. Giggenbach, W.F., 1988. Geothermal solute equilibria: derivation of Na–K–Mg–Ca geoindicators. Geochim. Cosmochim. Acta 52, 2749–2755. Goff, F., Bergfeld, D., Janik, C.J., Counce, D., Murrell, M., 2002. Geochemical data on waters, gases, scales, and rocks from the Dixie Valley Region, Nevada. In: Los Alamos National Laboratory Report LA-13972-MS, Los Alamos, NM, 71 pp.

L. Peiffer et al. / Geothermics 51 (2014) 154–169 Harrill, J.R., Hines, L.B., 1995. Estimated natural ground-water recharge, discharge, and budget for the Dixie Valley area, west-central Nevada. In: U.S. Geological Survey, Water-Resources Investigations Report 95-4052, 12 pp. Holland, T.J.B., Powell, R., 1998. An internally consistent thermodynamic dataset for phases of petrological interest. J. Met. Geol. 16, 309–343. Iovenitti, J.L., Blackwell, D.D., Sainsbury, J.S., Tibuleac, I.M., Waibel, A.F., Cladouhos, T.T., Karlin, R., Kennedy, B.M., Isaaks, E., Wannamaker, P.E., Clyne, M.T., Ibser, F.H., Callahan, O.C., 2012. Towards developing a calibrated EGS exploration methodology using the Dixie Valley geothermal system, Nevada. In: Proceedings, Thirty-Seventh Workshop on Geothermal Reservoir Engineering Stanford University, Stanford, CA, January 30–February 1, 15 pp. Johnson, J.W., Oelkers, E., Helgeson, H.C., 1992. SUPCRT92: a software package for calculating the standard molal thermodynamic properties of minerals, gases, aqueous species and reactions from 1 to 5000 bar and 0 to 1000 ◦ C. Comput. Geosci. 1992 (18), 899–947. Kennedy, B.M., Janik, C., Benoit, D., Shuster, D.L., 1999. Natural geochemical tracers for injectate fluids at Dixie Valley. In: Proceedings 24th Workshop on Geothermal Reservoir Engineering, Stanford Univ. Report SGP-TR-162. Kennedy, B.M., Van Soest, M.C., 2006. A Helium isotope perspective on the Dixie Valley, Nevada, hydrothermal system. Geothermics 35, 26–43. Lasaga, A.C., 1984. Chemical kinetics of water–rock interactions. J. Geophys. Res. 89, 4009–4025. Lutz, S., Moore, J., Benoit, D., 1997. Geologic framework of Jurassic reservoir rocks in the Dixie Valley geothermal field, Nevada: implications from hydrothermal alteration and stratigraphy. In: Proceedings 22nd Workshop on Geothermal Reservoir Engineering, Stanford Univ. Report SGP-TR-155, pp. 131–213. Lutz, S., Moore, J., Benoit, D., 1998. Integrated alteration mineralogy and fluidinclusion study at the Dixie Valley geothermal field, Nevada. In: Proceedings 23rd Workshop on Geothermal Reservoir Engineering, Stanford Univ. Report SGP-TR-IS8, pp. 315–332. McKenna, J.R., Blackwell, D.D., 2004. Numerical modeling of transient Basin and Range extensional geothermal systems. Geothermics 33, 457–476. Nimz, G., Janik, C., Goff, F., Dunlap, C., Huebner, M., Counce, D., Johnson, S., 1999. Regional hydrology of the Dixie Valley geothermal field, Nevada: preliminary interpretations of chemical and isotopic data. In: Lawrence Livermore National Laboratory Publication, UCRL-JC-135417, 9 pp. Palandri, J.L., Kharaka, Y.K., 2004. A compilation of rate parameters of water–mineral interaction kinetics for application to geochemical modeling. In: U.S. Geological Survey Open File Report, 1068, 64 pp. Pang, Z.H., Reed, M.H., 1998. Theoretical chemical thermometry on geothermal waters: problems and methods. Geochim. Cosmochim. Acta 62, 1083–1091. Phalen, W.C., 1919. Salt resources of the United States. U.S. Geol. Survey Bull. 669, 137–148.

169

Reed, M.H., Spycher, N.F., 1984. Calculation of pH and mineral equilibria in hydrothermal waters with application to geothermometry and studies of boiling and dilution. Geochim. Cosmochim. Acta 48, 1479–1492. Reed, M.H., Palandri, J., 2006. SOLTHERM.H06, A Database of Equilibrium Constants for Minerals and Aqueous Species. University of Oregon, Eugene, OR. Reheis, M., 1999. Extent of Pleistocene lakes in the Western Great Basin. In: U.S. Geological Survey Miscellaneous Field Investigation Map, MF-2323, 1 sheet. Sass, J.H., 1995. Mining the Earth’s heat in the Basin and Range. Geoth. Resour. Council Bull. 24, 125–129. Shevenell, L., De Rocher, T., 2005. Evaluation of chemical geothermometers for calculating reservoir temperatures at Nevada geothermal power plants. Geoth. Resour. Council Trans. 29, 303–308. Speed, R.C., 1976. Geologic map of the Humboldt Lopolith, Geological Society of America Map and Chart Series MC-14, 1:81050 Scale, 4 pp. Spycher, N., Peiffer, L., Saldi, G., Sonnenthal, E., Reed, M.H., Kennedy, B.M., 2014. Integrated multicomponent solute geothermometry. Geothermics 51, 113–123. Spycher, N., Sonnenthal, E., Kennedy, B.M., 2011. Integrating multicomponent chemical geothermometry with parameter estimation computations for geothermal exploration. Geoth. Resour. Council Trans. 35, 663–666. Thompson, G.A., Burke, D.B., 1973. Rate and direction of spreading in Dixie Valley, Basin and Range Province, Nevada. Geol. Sot. Am. Bull. 84, 627–632. Vanderburg, W.O., 1940. Reconnaissance of Mining Districts in Churchill County. U.S. Bur. Mines Inf. Circ. 7093, Nevada, 57 pp. Waibel, A.F., 1987. An overview of the geology and secondary mineralogy of the high temperature geothermal systems in Dixie Valley, Nevada. Geoth. Resour. Council Trans. 11, 479–486. Wanner, C., Peiffer, L., Sonnenthal, E., Spycher, N., Iovenitti, J., Kennedy, B.M., 2013. Assessing thermo-hydrodynamic–chemical processes at the Dixie Valley geothermal area: a reactive transport modeling approach. In: Proceedings 38th Workshop on Geothermal Reservoir Engineering, Stanford Univ. Report SGP-TR-198. Wanner, C., Peiffer, L., Sonnenthal, E., Spycher, N., Iovenitti, J., Kennedy, B.M., 2014. Reactive transport modeling of the Dixie Valley geothermal area: insights on flow and geothermometry. Geothermics, http://dx.doi.org/10.1016/j.geothermics.2013.12.003. Williams, C.F., Sass, J., Grubb, F., 1997. Thermal signature of subsurface fluid flow near the Dixie Valley geothermal field, Nevada. In: Proceedings 22nd Workshop on Geothermal Reservoir Engineering. Stanford Univ. Report SGP-TR-155, pp. 161–216. Xu, T., Spycher, N., Sonnenthal, E., Zhang, G., Zheng, L., Pruess, K., 2011. TOUGHREACT version 2.0: a simulator for subsurface reactive transport under non-isothermal multiphase flow conditions. Comput. Geosci. 37, 763–774.