• Open Access

Characteristics of CO2-driven cold-water geyser, Crystal Geyser in Utah: experimental observation and mechanism analyses


Corresponding author: Weon S. Han, Department of Geosciences, University of Wisconsin-Milwaukee, P.O. Box 413, Lapham Hall 366, Milwaukee, WI 53201, USA.

Email: hanw@uwm.edu. Tel: +1 414 229 2493. Fax: +1 414 229 5452.


Geologic carbon capture and storage (CCS) is an option for reducing CO2 emissions, but leakage to the surface is a risk factor. Natural CO2 reservoirs that erupt from abandoned oil and gas holes leak to the surface as spectacular cold geysers in the Colorado Plateau, United States. A better understanding of the mechanisms of CO2-driven cold-water geysers will provide valuable insight about the potential modes of leakage from engineered CCS sites. A notable example of a CO2-driven cold-water geyser is Crystal Geyser in central Utah. We investigated the fluid mechanics of this regularly erupting geyser by instrumenting its conduit with sensors and measuring pressure and temperature every 20 sec over a period of 17 days. Analyses of these measurements suggest that the timescale of a single-eruption cycle is composed of four successive eruption types with two recharge periods ranging from 30 to 40 h. Current eruption patterns exhibit a bimodal distribution, but these patterns evolved during past 80 years. The field observation suggests that the geyser's eruptions are regular and predictable and reflect pressure and temperature changes resulting from Joule–Thomson cooling and endothermic CO2 exsolution. The eruption interval between multiple small-scale eruptions is a direct indicator of the subsequent large-scale eruption.


Reduction in CO2 emissions has garnered overwhelming scientific support to ameliorate the impacts of global climate change (Kerr 2006; IPCC 2007). One method for mitigating CO2 concentrations in the atmosphere is to capture CO2 from its sources and directly sequester it into the deep subsurface. Deep saline formations, oil/gas reservoirs, and coal seams are potential sites for geologic CO2 storage (Bachu 2000; Orr 2004). However, gaps in our current knowledge still exist regarding the behavior of CO2 introduced into the subsurface environment (Gale 2004). Thus, comprehensive understanding of the dynamic subsurface processes associated with anthropogenic manipulations of CO2 is essential for successful geologic CO2 sequestration programs (DePaolo & Orr 2008).

One of the primary concerns is that stored CO2 could leak to the surface (IPCC 2005; Moore et al. 2005; Lewicki et al. 2007; Kano et al. 2009; Lu et al. 2010). Leakage mechanisms from CO2 storage sites include (i) fast-flow path leakage, which primarily involves CO2 movement through either transmissive faults or fractures preserved in the caprock (Chang et al. 2008; Pruess 2008a; Zhang et al. 2009), (ii) poorly sealed injection and monitoring well casings/cement (Bachu & Bennion 2009; Barlet-Gouedard et al. 2009; Wertz et al. 2009; Kampman et al. 2012), and (iii) improperly abandoned wellbores (Gasda et al. 2004; Nordbotten et al. 2004; Watson & Bachu 2009). The latter mechanism represents slow leakage, primarily involving CO2 transport by diffusional processes (Altevogt & Celia 2004; Klusman 2006; Annunziatellis et al. 2008). Diffusional CO2 leakage occurs without sensible signatures, and is difficult to detect without sophisticated sensing techniques, which include the use of CO2 flux chambers, eddy covariance towers, and remote sensing (Lewicki et al. 2003, 2011; Govindan et al. 2011). In some cases, changes in biological activities associated with CO2 emissions can be observed (Rogie et al. 2001; Farrar et al. 2002).

In geologic sequestration sites, captured CO2 is injected under supercritical conditions (>7.38 MPa and >31.1°C) (Span & Wagner 1996). Under these conditions, CO2 is highly compressible with compressibility factors ranging from 0.2 to 1 (Cole 2000). This results in a substantial change in CO2 volume and potentially inaccurate prediction of the stored CO2 plume size and geometry (Vilarrasa et al. 2010). As the compressed CO2 reaches pre-existing highly permeable faults and fractures that breach the caprock, leakage to other formations can occur (e.g., freshwater aquifers or the surface) (Shipton et al. 2005; Dockrill & Shipton 2010). The pressure gradient within these highly permeable conduits radically changes from the stored formation to the land surface. Assuming a fast-flowing path to the surface, CO2 escaping from the storage formation could instantaneously reach the land surface while experiencing the adiabatic expansion under isenthalpic conditions, which results in Joule–Thomson (JT) cooling (Pruess 2008b; Han et al. 2010). The temperature disequilibrium of CO2 due to pressure reduction under isenthalpic conditions can be estimated with the limiting ratio of temperature to pressure (Katz & Lee 1990) where the JT coefficient (μJT) is defined by the expression:

display math(1)

where P and T are pressure and temperature, respectively. The subscript H denotes the specific enthalpy and its derivative indicates that changes in P and T occur at constant specific enthalpy. The JT effect represents the temperature change under isenthalpic CO2 expansion. In the application to CO2 leakage problems, the temperature drop in CO2 due to pressure reduction under isenthalpic conditions has been studied by Paterson et al. (2008) and Pruess (2011). Fig. 1 represents the contours of μJT for CO2 calculated from the equation of state algorithms developed by Span & Wagner (1996). In this plot, the line where μJT is zero indicates the JT inversion curve. When μJT is positive, CO2 heats by compression and cools upon expansion.

Figure 1.

Contours of JT coefficient [°C MPa−1] in PT space calculated from Span & Wagner (1996). JT, Joule–Thomson.

Fast-flowing CO2 eruptions are primarily observed at petroleum fields where drilling operators must stop the uncontrolled CO2 emissions from hydrocarbon production wells (Lynch et al. 1985; Kouba et al. 1993; Skinner 2003). With ongoing geologic CO2 sequestration efforts, uncontrolled blowouts of CO2 injection wells become a major concern. Consequently, significant effort has been devoted to understanding the conditions and dynamics of well-blowout processes using computer simulations (Paterson et al. 2008; Pruess 2008b; Wertz et al. 2009) and field-oriented natural analogue studies (Lu et al. 2005; Gouveia & Friedmann 2006). The present study evaluates the CO2-driven cold-water eruption dynamics at a natural analogue site, Crystal Geyser in Utah, where CO2 and brine periodically erupt from an abandoned gas well. Other cold-water geysers of the world are listed in Table 1 (Glennon & Pfaff 2005).

Table 1. CO2-saturated cold-water geysers in the world, modified from Glennon & Pfaff (2005)
Crystal GeyserGreen River, Utah, USA15–20 m11–18 h15–45 min
Woodside GeyserWoodside, Utah, USA6–10 m28 min1.0–1.5 h
Champagne GeyserGreen River, Utah, USA7–8 m2 h5 min
Tenmile GeyserGreen River, Utah, USA2.5–3.5 m6 h 42 min51 sec
Tumbleweed GeyserGreen River, Utah, USA0.3–1.5 m2–8.5 min46–94 min
Unnamed GeyserSalton Sea, California, USA0.1–0.5 m10–60 secSec
Jones Fountain of LifeClearlake, California, USA<1.0 m60 min22 min
Cold Water GeyserYellowstone, Wyoming, USA0.5 mUnknown10 min
Source Intermittente de VesseBellerive, France1–6 m230–270 min45–50 min
Andernach GeyserAndernach, Germany40–60 m1.5–4 h7–8 min

Boiling Fount

Local name: Brubbel

Wallenborn, Germany2–3 m30 minA few minutes
Mokena GeyserNorth Island, New Zealand0.5–5 mMin–hSec–min
Povremeni GeyserSijarinska, Serbia20 m9 min2 min
Herlany GeyserHerlany, Slovakia20–30 m32–34 h30 min

The behavior of Crystal Geyser has been considered an appropriate analogue for both catastrophic and diffusive gaseous CO2 leakage from engineered CCS sites (Allis et al. 2005; Shipton et al. 2005; Gouveia & Friedmann 2006; Wilkinson et al. 2008; Heath et al. 2009; Kampman et al. 2012). The primary driving force controlling the periodic eruption of cold-water geysers is hypothesized to be the leakage and accumulation of natural CO2 within the well, but details of the eruption mechanisms are not understood comprehensively. Further detailed studies on natural CO2 leakage mechanisms will enhance the ability to appropriately monitor stored CO2 plumes and help to plan for public acceptance in engineered CO2 storage sites.

Site Descriptions

Crystal Geyser

In the Colorado Plateau region encompassing southeastern Utah and southwestern Colorado, natural CO2 reservoirs occur within the Woodside, Farnham, and McElmo domes (Fig. 2A) (Allis et al. 2001). In these reservoirs, over 98% pure CO2 has been trapped in dome-like structures and produced for the commercial usages such as dry ice, industrial uses, and enhanced oil recovery (Weeter & Halstead 1982; Gerling 1983; Roth 1983). However, sources and secondary migration of trapped CO2 at these reservoirs are poorly understood due to the multiple processes of their originations such as methanogenesis, biodegradation, kerogen decarboxylation, hydrocarbon oxidation, decarbonation of marine carbonates, and degassing of magmatic bodies (Selly 1998; Wycherley et al. 1999).

Figure 2.

(A) Geographic location of CO2 natural reservoirs in Utah and Colorado, (B) potentiometric surface map of Navajo Formation and the schematic saturation map of groundwater within the Navajo Formation (Hood & Patterson 1984). Arrows delineate the direction of groundwater movement, indicating that the areas of San Rafael Swell and Crystal Geyser are recharge and discharge zones, respectively. Stars (★) indicate the locations of CO2-saturated geysers and springs adjacent to Little Grand Wash and Salt Wash faults. (C) CO2 flux data surveyed during the field trips of August 4–6 and September 1–5, 2010.

Of interest to the study area, Hood & Patterson (1984) demonstrated that the San Rafael Swell and the Green River serve as recharge and discharge zones, respectively, for regional groundwater flow (Fig. 2B). Several cold-water geysers and springs listed in Table 1 are located adjacent to the Green River where two major east–west faults, the Little Grand Wash and the Salt Wash Graben faults, trend roughly parallel to each other (Shipton et al. 2005; Williams 2005; Heath et al. 2009; Kampman et al. 2009; Dockrill & Shipton 2010). Among these springs and geysers is Crystal Geyser, located immediately north of the Little Grand Wash fault and approximately 6 km south of the town of Green River, Utah (Barton & Fuhriman 1973; Baer & Rigby 1978; Waltham 2001). Anomalously high soil CO2 fluxes (>700 g m−2 day) have been measured along traverses perpendicular to these two fault zones (Allis et al. 2005). During August 4–6 and September 1–5, 2010, we measured additional CO2 flux adjacent to Crystal Geyser. The highest CO2 flux recorded was approximately 1600 g m−2 day (Fig. 2C).

Historical evolution of geyser eruption patterns

In July of 1869, the well-known geologist, John Wesley Powell, reported that some inactive mineral springs had created curious rock deposits (Powell 1987). These deposits attracted Glen M. Ruby who drilled a 40.64-cm-diameter exploration drilling hole to a depth of 800 m at the crest of a broad anticline between November 1935 and July 1936 (Kelsey 1991). The casing record is not available but presumably is <30 m deep (Barton & Fuhriman 1973). When the well was first drilled, the name of the geyser was ‘Ruby #1 State well’, and later Crystal Geyser became a popular name.

The patterns (uni- or bimodal) and duration of the eruptions at Crystal Geyser have evolved since the well had been drilled (Table 2 and Fig. 3). In 1936, the geyser bimodally erupted every 15 min with a height of 25 m and every 9 h with a height of 45 m (Kelsey 1991). Then, the eruption pattern changed from bimodal to unimodal between the 1930s and 1960s. The geyser erupted with one pattern, every 5 h for 5–10 min with a height of 60 m in the 1960s (Rinehart 1980). While Baer & Rigby (1978) and Barton & Fuhriman (1973) investigated a methodology to prevent 120 m3 per day of saline brine emitted from Crystal Geyser flowing into the Green River between 1968 and 1972, they observed that the geyser unimodally erupted every 4.25 h with a duration of 6–8 min (Fig. 3). Later, Martinez (1976) reported that the eruption intervals had increased to approximately 7 h by the 1970s. Based on the historical records, the eruption duration of the Crystal Geyser was consistently <10 min up to the 1970s.

Table 2. Historically observed eruption patterns in Crystal Geyser
DateTime of eruption beganEruption duration (min)Eruption intervalEruption height (m)Temperature (°C)References

Short 15 min

Long 9 h

Short 25

Long 45

 Kelsey (1991)
Late 1960s 5–10approximately 5 h6015Rinehart (1980)
06/13/196807:15 4 h 00 min21–2715.5–17.8Baer & Rigby (1978)
06/13/196811:156.84 h 15 min   
06/13/196815:307.34 h 15 min   
06/13/196820:007.04 h 15 min   
07/19/197202:486.94 h 15 min   
07/19/197208:257.74 h 15 min   
07/19/197213:246.14 h 15 min   
07/19/197218:277.24 h 15 min   
07/19/197223:416.94 h 15 min   
07/20/197205:10 4 h 15 min   
07/20/197210:407.54 h 15 min   
04/29/198916:5019 18–4020Murray (1989)
04/30/198906:301613 h 40 min   
04/30/198918:501912 h 20 min   
05/01/198908:531714 h 3 min   
08/13/198917:191116 h 1 min   
10/29/198904:192014 h 47 min   
10/29/198919:402015 h 21 min   
10/30/198911:342615 h 54 min   
10/31/198902:552015 h 21 min   
10/31/198916:392113 h 44 min   
11/01/198903:332015 h 54 min   
11/01/198922:512514 h 18 min   
1991  approximately 15 h25–30 Kelsey (1991)
1991    18Mayo et al. (1991)
     16Shipton et al. (2004)
06/10/2004 20   Glennon & Pfaff (2005)
     17.7Heath et al. (2009)
     18Kempman et al. (2009)
Figure 3.

Historical evolution patterns at Crystal Geyser indentified by eruption duration and the time until the start of the following eruption.

More recently, after observing 14 eruptions, Murray (1989) reported that the geyser erupted every 13–15 h with a duration of 16–25 min. The erupted water column reached a height of 18–40 m. Relative to the observation by Baer & Rigby (1978), both the eruption duration and intervals became approximately three times longer. Until 2004, the eruption duration remained relatively constant at approximately 20 min (Glennon & Pfaff 2005). Between August 2005 and November 2005, a comprehensive data set was collected by installing pressure and temperature sensors attached to the hole at the surface (Gouveia & Friedmann 2006). They were interested in studying the Crystal Geyser's eruptions because the CO2-driven geyser eruption mechanism could be analogous to potential catastrophic CO2 leakage from failed CO2 storage sites. Capturing 140 eruptions over 76 days revealed that the eruption patterns of the geyser returned from a unimodal to a bimodal eruption with longer eruption durations requiring more recharge time than shorter eruption durations (Fig. 3); especially, all historical observations before 2004 indicated that the eruption duration was <20 min. However, in 2005, the duration reached more than 2 h. Furthermore, our data set collected in 2010 showed that the eruption duration was significantly longer (B-type: 1 h and D-type: 5–7 h) than observed by Gouveia & Friedmann (2006).

Gas compositions and CO2 emission rates

Baer & Rigby (1978) reported that the emitted water from Crystal Geyser is always saturated with CO2, but they did not quantify the gas composition of the total discharge. A decade later, Mayo et al. (1991) analyzed the emitted gases discharged from Crystal Geyser, revealing that the gas composition is 96.15% CO2 with minor amounts of CH4. Heath et al. (2009) extended the survey to the adjacent geyser/springs shown in Fig. 2B and confirmed that all samples are gaseous CO2-rich (95.66–99.41%) with minor amounts of Ar, O2, and N2. Even though no H2S gas had been detected in their analyses, both Heath et al. (2009) and Shipton et al. (2005) reported the odor of H2S was in the vicinity of the geyser during eruptions. Throughout the field investigation, we also detected the strong scent of H2S gas.

In addition to the efforts quantifying the emitted gas compositions, Barton & Fuhriman (1973) and Baer & Rigby (1978) measured the amount of brine discharge and estimated that approximately 2730 metric tons of Na, Ca, K, and Mg salts are contributed to the Green River annually. Gouveia & Friedmann (2006) predicted the gaseous CO2 emission rate based on continuous measurements of pressure for 76 days. They calculated a CO2 emission rate of 2.6–5.8 kg sec−1 with a median of 5 kg sec−1. At this rate, Crystal Geyser produces 11 000 metric tons of CO2 per year.

Origin and source of the CO2

Baer & Rigby (1978) suggested that the CO2 gas emitted from Crystal Geyser would be a by-product of chemical reactions between ground water and Navajo Sandstone encountered at the depth of 215 m and hypothesized that the primary control of the geyser eruption was the exsolution of dissolved CO2, which regularly builds up enough pressure to cause the geyser eruptions. In their field experiments, the geyser eruptions became less frequent, and a smaller volume of water was emitted when the water in the geyser was intentionally pumped (e.g., decreasing the rate of the pressure build-up). In contrast, when the emitted water was pushed back into the well, the eruption was more frequent, and eventually, a large volume of water was emitted. Similar to the idea of Baer & Rigby (1978) and Assayag et al. (2009) hypothesized that free-CO2 gas is primarily originated from the exsolution of CO2-saturated brine in Navajo Sandstone. In order to test this hypothesis, they analyzed δ13C of both brine (−0.94 to −0.19 per mil) and free CO2 gas (−7.3 per mil) at Crystal Geyser and predicted that the CO2 degassing depth is relatively shallow (approximately 200 m).

Different from Baer & Rigby (1978) and Assayag et al. (2009), Mayo et al. (1991) hypothesized, based on the carbon isotope value (δ13C = −1.2 per mil) of math formula concentration in the fluid, that the CO2 gas originated from the thermal decomposition of carbonate rocks. Shipton et al. (2004) supported the thermal decomposition theory after analyzing carbon isotopes of travertine deposits and veins at Crystal Geyser, Tenmile Geyser, and Torrey's Spring where the values (δ13C = +4 to 5 per mil) were unusually enriched. At Crystal Geyser, the CO2 could have been produced by contact metamorphism of marine carbonates deposited below the Paradox Formation during the emplacement of Tertiary intrusions, some of which are now exposed in the La Sal and Henry Mountains. The thermal decomposition theory was also supported by Cappa & Rice (1995), who showed similar evidence that CO2 at McElmo Dome in southern Colorado originated from thermal decomposition of the Mississippian Leadville Limestone. Additionally, Shipton et al. (2004) suggested that CO2 could be originated from biologically mediated hydrocarbon generation due to the presence of oil seeps adjacent to Crystal Geyser.

Heath et al. (2009) extended the analyses to both helium (3He/4He = 0.3) and carbon isotopes (approximately −6.60 per mil) in gaseous CO2 and, based on these analyses, suggested that the CO2 originated at a depth of over 800 m as a free gas-phase CO2, as a product of the clay–carbonate reactions, the thermal degradation of carbonates, or a combination of these two processes. Further characteristics of CO2 sources were delineated by Wilkinson et al. (2008), who collected gas samples from Crystal Geyser and nearby geysers/springs shown in Fig. 2B and analyzed them for their helium, neon, argon, krypton, and stable carbon isotopes. Based on these analyses, Wilkinson et al. (2008) concluded that 1–20% by volume of the emitted CO2 originated from the mantle with the remainder derived from crustal sources. Exceptionally, Tenmile Geyser, which is located above the Salt Wash fault, showed a different origin for the CO2; 16–99% of CO2 is from the mantle origin with the remaining CO2 originating from the crust.

In summary, the conceptual models of Shipton et al. (2004) and Heath et al. (2009) imply the presence of a large natural CO2 reservoir beneath Crystal Geyser similar to ones found at McElmo Dome (Cappa & Rice 1995) and Springerville-St. John's, Arizona (Moore et al. 2005). In contrast, Wilkinson et al. (2008) argued that the noble gas composition of Ne, He, and CO2/3He ratio in Crystal Geyser and the adjacent geyser/springs is markedly distinct from large accumulations of natural CO2 reservoirs (Gilfillan et al. 2008). Furthermore, the size of the travertine deposits at Crystal Geyser and adjacent locations (hundreds of square meters) is much smaller than the travertine at Springerville-St. John's Dome where these deposits cover approximately 250 km2 (Moore et al. 2005; Dockrill & Shipton 2010). As discussed, the cause of the periodic eruptions and the origin of CO2 in Crystal Geyser are still debated. The goal of this study is to improve the current understanding by interpreting a new field-collected data set.

Experimental Design

We measured a time series change of pressure and temperature during three different periods to evaluate: (i) the present-day periodicity of the geyser eruptions; (ii) the role of pressure and temperature on the eruption periodicity and intensity; and finally, (iii) the historical evolution of the geyser's characteristics. We measured pressure and temperature every 20 sec at depths of 6.5 and 14 m in three periods from August 4 to 6, September 1 to 5, and December 8 to 16, 2010. During August 4–6, 2010, we determined the maximum depth where transducers could be installed and found an unknown obstacle at a depth of 16 m (Fig. 4). The nature of well blockage was not identified in this study.

Figure 4.

Schematic diagram of well configuration and transducer installation.

Originally, the geyser was drilled to a depth of 800 m in 1936. Barton & Fuhriman (1973) reported that the geyser was only open to 120 m. Additionally, a previous researcher who studied at Crystal Geyser 20 years ago indicated that he had no problem inserting a video camera deeper than tens of meters (A.L. Mayo, personal communication). During the field trip, the accessible depth of the geyser was much shallower (16 m). The exact history of any events clogging the well is unknown. However, anecdotal reports and discussions with several local people suggested that the hole was dynamited to enhance the eruption intensity of the adjacent geysers, and someone dumped railroad tiles for capping the geyser (Murray 1989; Shipton et al. 2004; Glennon & Pfaff 2005). In 2011, a digital camera was lowered to approximately 30 m in the hole, and it revealed that the hole was filled with gravel- to cobble-sized rocks (J.P. Evans, personal communication).

In order to acquire an in situ time series data set, several 2.5-cm-diameter PVC pipes were connected together and inserted into the geyser after attaching two Geokon Model 4500 vibrating-wire transducers. Pressure and temperature were measured every 20 sec at depths of 6.5 and 14 m (Fig. 4). The Geokon transducers utilize a pressure-sensitive stainless steel diaphragm to which a wire is connected. The other end of the wire is attached to a fixed position in the body of the instrument.

Eruption Dynamics at Crystal Geyser

Characteristics of a single eruption cycle

Figure 5 shows the representative time series changes in pressure and temperature measured at a depth of 14 m on December 8–12, which explicitly captures the disturbance of fluid pressure and temperature during multiple eruption periods. The observed pressure and temperature changes indicate that each single-eruption cycle (SEC) consists of four successive eruption types (A, B, C, and D) and two recharge periods (R1 and R2). The approximate timescale of a SEC was 30–40 h.

Figure 5.

Time series changes in pressure and temperature measured every 20 sec at depths of 14 m on December 8–12, 2010. P4 and P5 indicate times at which two of the photographs in Fig. 6 were taken.

In the SEC, an A-type eruption consisting of multiple small-scale eruptions continued approximately for 10–15 h. The individual small-scale eruption within A-type eruption occurred every 0.42 h (25.2 min), and its eruption duration was approximately 0.12 h (7.2 min). Immediately after the end of the A-type eruption, a single event of large-scale eruption (B-type) occurred, lasting 0.9–1.2 h. At the end of the B-type eruption, the groundwater level inside the well dropped to a depth of 2–3 m beneath the surface. Then, the subsequent recharge period (R1) lasted 4–5 h and preceded the C-type eruption, which consists of multiple events of small-scale eruptions similar to the A-type, but differing in that the total duration of C-type eruption is shorter (5–7 h). The final phase of the eruptive cycle is represented by a single event of large-scale eruption (D-type) lasting for approximately 5–7 h followed by a 10-h recharge period (R2).

Multiple events of small-scale eruptions (A- and C-types)

We observed that, prior to the small-scale multiple A- and C-type eruptions, CO2 bubbles were actively emitted within the well (P1 in Fig. 6), indicating that the CO2-saturated brine was exsolving CO2 gas at certain depths below. After 5–10 min, the small-scale eruption began, with the brine discharge reaching the top of the casing (P2 in Fig. 6). At this moment, the measured pressure reached the smallest value and temperature began to decrease. When the small-scale eruption activity stopped, the brine no longer bubbled, but only a certain amount of CO2 gas was emitted instantaneously at the casing top with characteristic hissing sounds (P3 in Fig. 6) indicating that the eruption event was over.

Figure 6.

Selected time series of pressure and temperature showing the regularity of A-type eruption. P1, P2, and P3 indicate times at which three of the photographs were taken.

The average reduction in pressure and temperature at a depth of 14 m during A- and C-type eruptions was approximately 0.018 MPa and 0.4°C, respectively (Fig. 6). With the following equation, it is possible to predict the volume fraction (α) of CO2 gas above the transducer when the A- and C-typed eruption occurred:

display math(2)

Here, the emitted water density (ρw), gravity acceleration (g), and depth (h) are 1000 kg m−3, 9.8 m2 sec−1, and 14 m, respectively. ΔP is the average reduction in pressure (0.018 MPa) shown in Fig. 6. The density of CO2 (math formula) is predicted to be 2.4 kg m−3 at 0.13 MPa and 17.2°C. When the A- and C-type eruptions occurred, CO2 volume fraction (α) above the transducer was approximately 13%.

Comparison with earlier studies indicates that A- and C-type eruption patterns have evolved over time during the past several decades. Three previous studies reported the pattern of overflows, which presumably meant the small-scale A- and C- type eruptions defined in this study. Based on the visual observation by Martinez (1976), overflows were at hourly interval and lasted 1–2 min during 1960 and 1970. Murray (1989) reported that the overflowing pattern changed during the 1980s; its interval is shortened from 54 to 15 min, but the duration increased from 5 to 10 min. Later, Glennon & Pfaff (2005) observed the consistent pattern where the geyser overflowed every 30 min and lasted 5–10 min. The latter pattern is maintained throughout the present study.

Based on our measurements, the mean duration of the A- and C-type eruptions was consistent such that the mean and standard deviation were 0.12 h (7.2 min) and 0.01, respectively, with approximately 30-min intervals between eruptions (Fig. 6). Although the intensity (consistent pressure reduction: 0.018 MPa) and duration of the individual small-scale eruptions remained relatively constant, we found that the interval between the eruptions decreased in proximity to the single large-scale B- and D-type eruptions (Fig. 7). The interval between the A- and C-type eruptions decreased from above 0.7 h (42 min) to approximately 0.3 h (18 min) as the large-scale B- and D-type eruptions became imminent. This pattern has never been observed in any studies related to hot- or cold-water geyser activity.

Figure 7.

Time intervals between A- and C-type eruptions prior to B- and D-type eruptions.

Previous studies of hot-water geyser eruptions revealed that the interval between eruptions could be correlated in part with external forces including barometric pressure, tidal forces, and seismicity (Rinehart 1972, 1980; Hutchinson 1985; Silver & Vallette-Silver 1992). The monitoring period in this study is too short to confirm that the external forces could control the interval between two successive eruptions similar to the studies of hot-water geysers. Nevertheless, the in situ changes, triggered by exsolution of CO2 gas and subsequent reduction in hydrostatic pressure, appear to reduce the time interval between multiple small-scale eruptions. With this known trend, the interval between A- and C-type eruptions could serve as a predictor of B- and D-type eruptions at Crystal Geyser.

Single event of large-scale eruptions (B- and D-types)

The single event of B- and D-type eruptions is characterized by relatively large reductions in pressure (0.045 MPa) and temperature (0.6°C) and longer eruption durations (Fig. 5). The maximum height of the water column in these large-scale eruptions reached approximately 10 m above the land surface (P4 in Fig. 6). Immediately after the large-scale events began, the water column reached its maximum height reflecting to the lowest pressure observed in Fig. 5. The intensity then decreased until the arrival of the recharge period. Although the maximum height of the water column was similar in both B- and D-type eruptions, the duration was significantly different, representing approximately 1.2 and 4.7–6.8 h, respectively.

Both the B- and D-type eruptions can be broken down into five definite segments (Fig. 8A,B). The first segment indicates the small-scale eruptions such as A- and C-type eruptions. The pressure builds half way and is disrupted by a sudden increase in gaseous CO2 causing the sudden drop in pressure to a minimum value for the initiation of B- and D-type eruption cycles (segment II). As the eruption progresses, the pressure slowly builds and oscillates (segment III). The typical increase in pressure for the B-type eruption is within 0.01–0.03 MPa, while the D-type eruption only increases in pressure at a maximum of 0.01 MPa. Segment IV reflects a sharp increase in pressure, which is likely caused by a combination of the CO2 supply ceasing, while the water pool at the surface flows back into the well (P5 in Fig. 6). The final segment V is a sharp decrease in pressure, which occurs due to the sudden decrease in water level below 2–3 m within the well. Similar to our observation, previous studies also reported that the water pool surrounding Crystal Geyser was completely drained after the major eruption (Murray 1989; Glennon & Pfaff 2005). After B- and D-type eruptions, the recharge period began and lasted 3 and 10 h, respectively.

Figure 8.

Magnified view of representative B- and D-type eruptions, consistently composed of five segments during the eruption; (A) B-type eruption and (B) D-type eruption.

Interpretation of unique patters in pressure and temperature

Five specific patterns (I, II, III, IV, and V) were identified within the eruption cycles shown in Fig. 5. Patterns I and II represent the systematic coupled nature of pressure and temperature probably induced by JT cooling and endothermic CO2 exsolution. At Crystal Geyser, CO2 gas exsolved from the CO2-saturated brine at a certain depth below the surface (Baer & Rigby 1978; Assayag et al. 2009). Immediately after the exsolution of CO2 gas bubbles, the CO2 gas migrates vertically through the water column due to the buoyancy. As CO2 gas migrates closer to the surface, the hydrostatic pressure becomes smaller and the volume of CO2 gas increases more. If these processes occur instantaneously, a substantial decrease in temperature will occur (Oldenburg 2007; Han et al. 2010). The JT effect addresses the temperature drop in CO2 due to the instantaneous expansion of CO2 induced by pressure reduction under isenthalpic conditions (Katz & Lee 1990). For example, the single event of B- and D-type eruptions is characterized by relatively large reductions in both pressure (approximately 0.045 MPa) and temperature (0.6°C). Immediately before a D-type eruption began, the detected pressure and temperature were 0.142 MPa and 17.2°C, respectively (Fig. 5), and the corresponding JT coefficient predicted from Span & Wagner (1996) is 11.684°C MPa−1 at these conditions. Once the eruption was initiated, the pressure is instantaneously reduced and reached the minimum, 0.09 MPa, resulting in an instantaneous expansion of CO2 gas. With Eq. (1), a pressure drop of 0.052 MPa (0.142 MPa minus 0.09 MPa) results in the 0.61°C decrease in temperature, which is close to our field observation shown in Fig. 5.

At a certain depth where CO2 bubbles nucleate from CO2-saturated brine, the temperature will be reduced due to endothermic CO2 exsolution. It is well understood that CO2 dissolution is an exothermic reaction with the appropriate range of enthalpy from −440 to −200 kJ kg−1 at temperatures from 25 to 100°C (Carroll et al. 1991; Duan & Sun 2003; Han et al. 2010). CO2 exsolution from liquid is the reverse process of CO2 dissolution in liquid. In principle, the corresponding enthalpy change in CO2 exsolution is going to be the same as CO2 dissolution except it is an endothermic reaction; brine cools as CO2 gas exsolves from the fluid. Han et al. (2012) simulated the formation temperature increase due to CO2 dissolution into a brine. At reservoir conditions of approximately 7 MPa and 41.7°C, the increase in the formation temperature due to CO2 dissolution is approximately 0.5°C. Similarly, even if the formation conditions are different at Crystal Geyser, the range of temperature reduction is predicted to be within a range of ±0.5°C.

At the field observation shown in patterns I and II of Fig. 5, large-scale eruptions caused greater reductions in pressure than small-scale eruptions, resulting in a greater temperature decrease due to JT cooling. A magnified view of pattern I clearly illustrates three consistent and recurring events: pre-eruption, eruption, and posteruption (Fig. 6). During the pre-eruption period, pressure was relatively constant, but a 0.2-C increase in temperature was observed. The cause of this increase is interpreted to be the recharge of relatively warm groundwater from deeper formations. During the eruption period lasting 0.12 h (7.2 min), both pressure and temperature dropped rapidly. The reduction in hydrostatic pressure indicates that a certain portion of the water column above the transducer was displaced by CO2 bubbles, which were nucleated at deeper depths and migrated vertically due to buoyancy. As noted above, the temperature drop is primarily caused by both the JT effect on CO2 and endothermic CO2 exsolution. After the cessation of eruptive activity, the temperature again increases due to heat conduction with the surroundings and recharge by relatively hot groundwater.

Pattern III in Fig. 5 is restricted to the end of B- and D-type eruptions. At this point, the measured temperature dropped to a minimum, but the pressure increased suddenly. As the large-scale eruptions continued, typically over a period of 1–5 h, fluid in the wellbore is cooled due to continued CO2 exsolution and adiabatic expansion. Pattern IV shows the lowest pressure at the beginning of B- and D-type eruptions, when the erupting water column reached its maximum height. Finally, we observed the slight increase in temperature immediately after the end of recharge period (pattern V in Fig. 5). The slight increase in temperature is interpreted to be due to recharge of relatively hot groundwater from deeper formations.

The complete data set monitored at depths of 6.5 and 14 m during September 1–5 and December 8–16, 2010 are shown in Fig. 9. The previously defined SEC and repetition of the five patterns shown in Fig. 5 are evident. However, the temperature perturbation effect was predominant at deeper depths.

Figure 9.

Time series changes in pressure and temperature measured every 20 sec at depths of 6.5 and 14 m; (A) September 1–5, 2010 (*during the D-type large-scale eruption between 13:10 pm September 2 and 1:10 am on September 3, the PVC pipe holding the transducers is broken due to the hydraulic power of eruptions. Pipe with the transducers fell down but stayed still above an unknown obstacle shown in Fig. 4. Since the transducers are relocated from the designated depth, the measured pressures at both 6.5 and 14 m are slightly increased and maintained the same level of pressures during the rest of observation period) and (B) December 8 to 16, 2010.

Conceptual Description of Crystal Geyser Eruptions

The conceptual description of Crystal Geyser eruptions discussed here is based in part on the study of Lu et al. (2005), who measured pressure and temperature variations at the CO2-driven cold-water geyser in Te Aroha, New Zealand. Due to the anthropogenic activities (e.g., dynamite, capping activity, etc.) at Crystal Geyser, the well presumably collapsed at certain depths, and thus, CO2 bubbles need to migrate through porous materials from nucleated depth to the surface.

The conceptual and experimental studies of hot geysers suggested that geysers should consist of a chamber in order to initiate periodic eruption activity. The chamber where the recharged cold water from shallow formation is boiled is necessarily placed at the lowest level of the well (White 1967; Steinberg et al. 1981; Manga & Brodsky 2006; Hutchinson et al. 2010). Similarly, it is presumed that, when Crystal Geyser was drilled to 800 m in 1936, the well was unintentionally connected to the pre-existing chamber and created the place where CO2-saturated brine is continuously recharged from the Navajo Sandstone approximately 200 m below.

In the conceptual model, stage A represents the start of the geysering cycle, when the CO2-saturated brine reaches the top of the well (Fig. 10). The hydrostatic pressure profile within the well is shown as Line A. The critical CO2 exsolution pressure point (Pex) indicates the pressure in the well, where CO2 starts to exsolve if the pressure is less than Pex. The critical depth, DA, is the corresponding depth at Pex. Presumably, CO2 will stay as a dissolved species in brine at the depth below DA, but CO2 will begin to exsolve above DA. Once CO2 exsolution is initiated, CO2 gas will instantaneously migrate to the surface and escape to the atmosphere. In stage A, both temperature and pressure at the observation point (★) are at their maximum values.

Figure 10.

Conceptual diagram describing the CO2-driven cold-water geyser, and the stars (★) indicate pressure and temperature at observation points in each stage.

In stage B, the refilling of the well at the bottom by the recharging aquifer causes the water level to rise continuously within the well. When the overflow starts, the few CO2 bubbles rising with the brine initiate the decrease in pressure profile from stage A to stage B. The water level is fixed at the well head, so any slight density change progresses downward the well as additional gaseous CO2 is released from the brine. The CO2 bubble growth rate is increasing, and more CO2 bubbles are coalescing. In stage B, the pressure and temperature at the observation points (★) are reduced relative to stage A. The DB corresponding to Pex becomes deeper as more CO2 bubbles are exsolved from the brine, resulting in the continued decrease in the pressure profile (stage B to stage C). At stage C, the Pex reaches its deepest depth (DC), and the primary geyser eruption begins. Then, the smallest hydrostatic pressure is reached in the well. Due to extensive CO2 bubble nucleation and adiabatic expansion (JT effect), the temperature at the observation points (★) reaches its lowest value.

Finally, once the eruption is over at stage D, the majority of the CO2 bubbles escape to the atmosphere, and the significant amount of brine discharges to the surface, resulting in a 2- to 3-m drop of the water level within the well (P5 of Fig. 6). Then, the depth (DD) of Pex becomes shallow. Recharge of CO2-saturated brine from the Navajo Sandstone continuously occurs. Once the groundwater level reaches the well tops, stage A begins again.

Changes in Eruption Characteristics due to Adjacent Seismic Activities

The eruption characteristics at Crystal Geyser (interval, duration, and eruption height) have changed significantly during the past 80 years (Fig. 3 and Table 2). Unfortunately, no other data has been collected to identify the influence of external forces on the geysering patterns in other CO2-driven cold-water geysers around world. However, several researchers have observed changes in the eruption patterns of thermally driven hot-water geysers (e.g., Old Faithful in Yellowstone, Wyoming and Steamboat Spring, Nevada). White (1967) reported that the average rate of discharge of Geyser 23-n at Steamboat Springs in Nevada is affected by barometric pressure; at low atmospheric pressure, discharge is increased. Rinehart (1972) extended the analyses and concluded that the geyser eruption is not only influenced by barometric pressure, but also by earth tidal forces and tectonic strain associated with earthquakes. However, the barometric and earth tide effect on geysers are controversial (White 1972). Recently, Rojstaczer et al. (2003) analyzed the historic eruption from six natural geysers at the upper basin in Yellowstone National Park and concluded that earth tide and diurnal pressure influences are not identifiable. Nevertheless, it seems that some geysers are influenced by distant earthquakes that generate seismic strains (Hutchinson 1985; Silver & Vallette-Silver 1992). The geyser interval in Yellowstone National Park changed following the 2002 M 7.9 Denali earthquake in Alaska despite the large distance of 3100 km separating the two (Husen et al. 2004). The arrival of large-amplitude surface waves could alter local permeability by opening or clogging existing faults or fractures. Indeed, Ingebritsen & Rojstaczer (1996) numerically modeled thermally driven periodic geysers and found that the geyser's periodicity is strongly dependent on the permeability contrast between the geyser chamber and the surrounding rock.

Similarly, we presume that the nearby seismic events could alter the Little Grand Wash fault system and disturb geyser eruption. Additionally, the presence of inactive travertine deposits from the Pleistocene to the present time along the Little Grand Wash fault zone suggests processes leading to self-sealing, or permeability enhancement has occurred (Dockrill & Shipton 2010). At the Crystal Geyser, we discovered that the eruption duration was significantly extended from 2 to 5–7 h between the time of the most recent two measurements (2006 and 2010) (Fig. 3). At the same period, we discovered that two seismic events with magnitudes of 1.17 and 2.63 occurred at hypocenters within 3 km of Crystal Geyser (Fig. 11). We hypothesize that these two events could have initiated movement of the Little Grand Wash fault system and disturbed the eruption patterns of Crystal Geyser. In addition, between 1965 and 1975, many seismic events with magnitude >2.0 were recorded adjacent to Crystal Geyser (Fig. 11). Presumably, these multiple seismic events could have induced the alteration in eruption intervals of Crystal Geyser. One implication of this critical observation is that the known cold-water eruptive systems are likely very sensitive to tectonic stresses. If so, it should be noted that engineered sequestration sites could be altered by such tectonic impacts.

Figure 11.

Seismic events occurring within 40 km of Crystal Geyser.


Crystal Geyser continuously leaks stored CO2 through a diffusive gas phase through local fault systems. However, the main leakage is episodic in nature, in the form of eruptions from the borehole that serves as the main conduit. The primary purpose of this study is to understand the eruption characteristics in Crystal Geyser including the present-day periodicity of the geyser eruptions, the coupled nature of pressure and temperature on the eruption periodicity and intensity, and finally, the historical evolution of the geyser's characteristics. The in situ measurements of pressure and temperature demonstrate that the eruptions occur regularly and are predictable to some extent; four successive eruptions (A, B, C, and D) and two recharge periods (R1 and R2) were documented (Fig. 5).

The mean duration of the A- and C-type eruptions was consistently 0.12 h (7.2 min), and their intervals were 30 min (Fig. 3). The average reduction in pressure and temperature at a depth of 14 m was approximately 0.018 MPa and 0.4°C, respectively (Fig. 6), and the predicted CO2 volume fraction during the eruptions was approximately 13%. Although the eruption intensity and duration remained relatively constant, the interval between the eruptions decreased in proximity to the single large-scale B- and D-type eruptions (Fig. 7). This pattern has never been observed in any previous studies related to hot- or cold-water geyser activity. The single event of B- and D-type eruptions is characterized by large reductions in pressure (0.045 MPa) and temperature (0.6°C) and longer eruption durations (B-type: 1.2 h and D-type: 4.7–6.8 h). These eruptions could be broken down into five definite segments (Fig. 8A,B).

In the observation of all eruption patterns, pressure and temperature perturbations were systematically coupled, induced by JT cooling on CO2, endothermic CO2 exsolution, and thermal conduction between CO2, water, and surrounding rocks. In addition, historical observations demonstrate that the eruption patterns have changed over the past eight decades (Fig. 3). Although the reasons for these changes are not yet established, we suggest that tectonic activity (Fig. 11) and precipitation and dissolution of calcium carbonate could have altered the fluid and gas flow paths. The direct application of these addressed findings is the prediction of potential modes of CO2 leakage from engineered sequestration sites as it relates to sustainable CO2 storage and the development of monitoring techniques (Holloway et al. 2007; Lewicki et al. 2007; Fessenden et al. 2009; Jeandel et al. 2010).


The authors would like to thank an anonymous reviewer and Jim Evans for their technical review and Amir Mijatovic and Rich Esser for assisting field work. All financial support for this research was provided by both National Science Foundation (EAR-1246404) and Korea National Oil Corporation funded by the Korea Institute of Energy Technology Evaluation and Planning (2011T100100331).