We investigate geologic controls on circulation in the shallow hydrothermal system of Masaya volcano, Nicaragua, and their relationship to surface diffuse degassing. On a local scale (∼250 m), relatively impermeable normal faults dipping at ∼60° control the flowpath of water vapor and other gases in the vadose zone. These shallow normal faults are identified by modeling of a NE-SW trending magnetic anomaly of up to 2300 nT that corresponds to a topographic offset. Elevated SP and CO2 to the NW of the faults and an absence of CO2 to the SE suggest that these faults are barriers to flow. TOUGH2 numerical models of fluid circulation show enhanced flow through the footwalls of the faults, and corresponding increased mass flow and temperature at the surface (diffuse degassing zones). On a larger scale, TOUGH2 modeling suggests that groundwater convection may be occurring in a 3–4 km radial fracture zone transecting the entire flank of the volcano. Hot water rising uniformly into the base of the model at 1 × 10−5 kg/m2s results in convection that focuses heat and fluid and can explain the three distinct diffuse degassing zones distributed along the fracture. Our data and models suggest that the unusually active surface degassing zones at Masaya volcano can result purely from uniform heat and fluid flux at depth that is complicated by groundwater convection and permeability variations in the upper few km. Therefore isolating the effects of subsurface geology is vital when trying to interpret diffuse degassing in light of volcanic activity.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Surface diffuse degassing is a direct result of magma decompression, and therefore changes in diffuse degassing can be related to changes in volcanic activity [Chiodini et al., 1998]. However, the connection between surface degassing and the magmatic source region often involves groundwater, which can strongly affect the location and magnitude of surface emanations that we can detect [Todesco, 2008]. Volcanic terrains in particular are also characterized by faults, fractures and permeability variations that strongly affect fluid flow paths and rates [Todesco, 1997; Caine et al., 1996; Evans et al., 2001; Manzocchi et al., 2008]. All of these features can change over time. Geophysical investigations can help to constrain these subsurface variations. Numerical simulations allow us to explore how these structures can affect fluid flow, and aid in relating surface diffuse degassing variations directly to their volcanic source.
 Volcanic activity perturbs groundwater flow by adding heat and gases to groundwater systems, manifest at the surface by flank degassing, hot springs, and flank fumaroles. Measurements of diffuse degassing have been applied as a volcano monitoring tool with varying degrees of success at Mt. Merapi, Indonesia [Toutain et al., 2009]; Mt. Fuji, Japan [Notsu et al., 2006]; Mt. Etna, Italy [Badalamenti et al., 2004]; the Phlegrean Fields, Italy [Granieri et al., 2003]; Stromboli volcano, Italy [Carapezza and Federico, 2000]; Taal volcano, Philippines [Zlotnicki et al., 2009]; Ruapehu volcano, New Zealand [Werner et al., 2006]; Nisyros, Greece [Brombach et al., 2001]; Mammoth Mountain, USA [Rogie et al., 2001]; and San Vicente volcano, El Salvador [Salazar et al., 2002]. These studies all indicate that a good geological model of the subsurface is a vital prerequisite for interpreting changes in diffuse degassing in light of volcanic activity. In this study, therefore, we performed detailed geophysical investigations of the structure and fluid circulation on the flank of Masaya cone at Masaya volcano, Nicaragua, to constrain numerical models of fluid transport.
 We collected magnetic data in a 133 m × 125 m zone of low-temperature diffuse degassing on the flank of Masaya cone, and used these data to infer local geological structures and permeability variations. From these, self-potential (SP) and CO2 flux distributions, and information about the hydrologic system gained from transient electromagnetic soundings [MacNeil et al., 2007], we created a TOUGH2 numerical simulation. The fluid flux output from the TOUGH2 model was then compared with CO2 flux measurements to refine a structural and hydrothermal model of the single flank degassing zone. On a larger scale, the consistent spatial distribution of the diffuse degassing zones was used to constrain a TOUGH2 hydrothermal model of the entire fracture zone along the flank of the cone.
2. Methods for Delineating Flow Paths
 The geophysical techniques that we employ in this paper are each sensitive to a different rock or fluid property, and each technique provides unique information about the system. Magnetic data are particularly useful for detecting structures in volcanic terrains, due to the high contrast in magnetic properties between basaltic lava, scoria, and alluvium [Stamatakos et al., 1997]. Magnetic profiles and maps have been used successfully to infer geological features such as faults at a number of sites [e.g., Jones-Cecil, 1995; Connor et al., 1997; La Femina et al., 2002], even in granitic areas where the magnetic contrast is relatively small [McPhee et al., 2004]. On basaltic volcanoes, magnetic anomalies associated with faults are often on the order of 100–1000 nT, two to three orders of magnitude larger than the magnetic anomalies induced through electrokinetic effects associated with fluid flow [Zlotnicki and Le Mouel, 1990; Adler et al., 1999].
 SP variations on volcanoes result primarily from fluid flow, and therefore provide information about fluid flow paths and rates. Three different relevant mechanisms generate SP anomalies [Bedrosian et al., 2007]: thermoelectric, electrokinetic and fluid-disruption effects. Although each can play a part, electrokinetic effects are theoretically larger [Corwin and Hoover, 1979] and most likely to dominate in a low-temperature system like the flank degassing areas at Masaya. Electrokinetic interaction between moving pore fluid and the electric double layer at the fluid–solid interface generates an electric potential gradient that is detectable with SP electrodes [Overbeek, 1952]. In extensional tectonic environments volcanic gas and vapor often rise buoyantly along faults [Goff and Janik, 2000], resulting in positive SP anomalies [Nishida et al., 1996; Michel and Zlotnicki, 1998; Finizola et al., 2002, Hase et al., 2005]. Negative SP anomalies around fissures and craters are interpreted as meteoric water recharge [e.g., Sasai et al., 1997].
 Current activity is limited to two cones (Nindiri and Masaya) and their small pit craters which include Masaya, San Pedro, Santiago, and Nindiri craters (Figure 1). There have only been two lava flows in the last 500 years; in 1670 a 1-km wide lava lake in Nindiri crater overflowed, and in 1772 lava was emitted from a fissure on the flank of Masaya cone [Rymer et al., 1998]. Santiago crater has become the main site of activity since its formation in ∼1858 [Rymer et al., 1998] and has been the site of very large noneruptive gas flux in recent decades [Stoiber et al., 1986; Horrocks et al., 1999; Burton et al., 2000; Delmelle et al., 2002; Duffell et al., 2003]. Crater gas flux and composition are consistent over time [Horrocks et al., 1999] and imply a magma body of approximately 10 km3 [Walker et al., 1993]. This is fed by gas-rich magma that ascends from depth, degasses, and sinks, producing an open, relatively stable system [Stix, 2007]. Since Masaya caldera formed 2–6 ka, renewed basaltic volcanism has also led to the formation of a series of cinder cones, fissure vents and lava flows within the caldera [Walker et al., 1993; Rymer et al., 1998; Williams-Jones et al., 2003]. During the past few years, however, changes in activity have been limited to vent-clearing ash explosions, lava ponding and rockfalls within Santiago crater (Figure 2).
 On the northeast flank of the volcano, outside of the pit crater network, there is a fault and fracture zone extending some kilometers where water vapor and CO2 are emitted at moderate temperatures (40–80°C) [Lewicki et al., 2003; Pearson et al., 2008; Mauri et al., 2012]. This radial fracture zone extends from Masaya cone past Comalito cinder cone (Figure 1b), where CO2 fluxes of up to 5.8 × 10−4 kg/m2s have been measured [Lewicki et al., 2003]. The fracture zone is approximately 100 m wide, as suggested by magnetic profiles, positive SP anomalies, and elevated CO2 flux. In some areas along its length it also forms a topographic offset at the surface.
 Within the caldera, the groundwater budget is controlled by the balance between meteoric recharge, evapotranspiration, and vent and diffuse degassing, which occur primarily in Santiago crater. This budget, supported by numerical models of groundwater transport, indicates that groundwater flow in the area of Comalito and the radial fracture zone is toward the active crater at depth, which acts as a groundwater sink due to vaporization [MacNeil et al., 2007]. Our study is primarily concerned with modeling the hydrothermal conditions that give rise to the flank zones of diffuse degassing.
4. Data Collection and Results
 Geophysical studies of the flank diffuse degassing area were completed over several years of field work (2003–2007). During this time there were no discernible changes in the location or nature of flank degassing, and volcanic unrest was limited to lava ponding and crater collapse in Santiago crater (Figure 2). Continuous monitoring showed that soil temperature, SP and CO2 flux were also relatively stable [Lehto, 2007; Pearson et al., 2008]. All measurements were recorded during the dry season, on dry days. A total of 4508 ground magnetic measurements were gathered at the middle degassing zone on the flank of the volcano in August 2007 (Figures 1 and 3a). They were recorded along approximately parallel transects covering an area of 133 m × 125 m. The readings were taken with a Geometrics, Inc. Portable Cesium Magnetometer G858, which has a sensitivity of 0.01 nT. Measurement locations were recorded using a GPS (Leica Geosystems Inc. GS20 Professional Data Mapper) with an accuracy of approximately 40 cm, using a GPS base station to improve accuracy (baseline < 1 km).
 Three SP and CO2 flux profiles were completed in May 2006. The profiles were approximately parallel and between 95 and 105 m long (Figure 3a), depending on topography and dimensions of the degassing zone. SP measurements were referenced to an electrode at the end of each line outside of the degassing area and were recorded every 1 m using non-polarizing Pb-PbCl2 electrodes and a high impedance voltmeter. CO2flux measurements were made every 2 m using a LI-COR, Inc. Li-800 portable gas fluxmeter [e.g.,Chiodini et al., 1998], which has an error of ±6% and a range of 0–20000 ppm. Atmospheric temperature and pressure remained stable throughout the day. An SP map was also made from data collected in 2003 (Figure 3c) and used to interpolate five transects covering the same area as the magnetic measurements.
 Magnetic measurements ranged between 32000 and 41000 nT, with an average of 35900 nT (Figure 3b and Table 1). This is consistent with the International Geomagnetic Reference Field (35400 nT, 37° inclination, 0.5° declination). No magnetic storms occurred during the sampling period, and total diurnal drift in the magnetic field at this latitude was <20 nT [Kerridge, 2001], negligible compared to mapped anomalies. There is a NE-SW trending magnetic anomaly of 2300 nT in the eastern portion of the map area. This anomaly corresponds remarkably well to the location and trend of the topographic offset, and is interpreted as a fault. The topographic offset and magnitude of the magnetic anomaly become significantly larger downslope within the zone of diffuse degassing. The magnetic anomaly amplitude ranges from 1000 nT along profile d-d′ to 2300 nT at profile a-a′ (Figure 3b and Table 1).
Table 1. Statistical Results From Magnetic, CO2 and SP Surveysa
SD stands for standard deviation. Anomalies are calculated using the nearby maxima and minima and are therefore sometimes less than the maximum difference along a profile.
Obs Mag (nT)
Calc Mag (nT)
2.1 × 10−5
2.8 × 10−6
4.9 × 10−6
2.1 × 10−5
2.2 × 10−5
4.3 × 10−6
6.1 × 10−6
2.2 × 10−5
2.6 × 10−5
2.2 × 10−8
3.6 × 10−6
5.2 × 10−6
2.6 × 10−5
 SP measurements ranged between −151 and 160 mV, with peak-to-peak amplitudes of 135 to 210 mV along a single profile. There is a NE-SW trending, 30-m-wide positive anomaly of up to 140 mV along the center of the map (Figure 3c). The anomaly decreases gradually to the west, more abruptly to the east (Figure 4). The eastern boundary of the anomaly reaches a minimum of −150 mV, and coincides with the topographic offset. Several of the SP profiles collected in 2006 show a gradual increase and then a more rapid decrease from NW to SE (Figure 5), also suggesting a NE-SW trending anomaly. The changes are concordant with fault locations estimated from topography and magnetic data. The total variation in SP during the map measurements in 2003 and the profile measurements in 2006 remained approximately 200 mV.
 CO2 flux measurements varied between 0 and 2.6 × 10−5 kg/m2s (Figure 5 and Table 1). Although Line 2 showed some agreement with SP data, CO2-flux variations were generally more abrupt and less consistent than changes in SP. Elevated CO2 flux was detected only NW of the topographic offset; there was a complete absence of CO2 to the SE (Figure 5).
5. Modeling and Interpretation
5.1. Magnetic Data
 The distribution and dip of faults in the study area were estimated from ground-magnetic data using Geosoft Inc. Oasis Montaj software. The total intensity of magnetization is the sum of the intensity of induced and remnant magnetization [Plouff, 1975]. We assumed constant remnant magnetization because all of the potential magnetic sources at Masaya are much younger than the last reversal of the Earth's magnetic field (∼0.78 Ma) [Cande and Kent, 1995], and tectonic rotations are thought to be minimal during this time. Therefore we were able to isolate the induced magnetization, creating a model using bulk values as we are only interested in the total vector of magnetization.
 A peak-to-peak amplitude of over 1000 nT (Figure 3b and Table 1) is consistent with vertical separation of basalt that has a strong magnetization [e.g., La Femina et al., 2002]. We consider the magnetic source to be a lava flow unit of high bulk magnetization beneath scoria of low bulk magnetization with some faulting. Other potential sources of the magnetic anomaly, such as a dike, would produce anomalies with a vector of magnetization inconsistent with those observed. Modeling is used to estimate the depth to the top of the lava beneath a veneer of scoria. We used vectors of magnetization of magnitude 2 × 10−2 for the lava flow and 1 × 10−6 for the scoria. This produces a plausible lava flow structure with a depth <20 m and a variable depth to the top of the lava of +/−10 m (Figure 4). This structural model and the observed topographic offset indicate the presence of SE-dipping normal faults in the map area.
5.2. Hydrothermal Models
 Numerical modeling of the hydrothermal system was done using the Petrasim interface to the TOUGH2 code. TOUGH2 is a sophisticated program to model multiphase and multicomponent fluid flow, evolved from the MULKOM code [Pruess, 2004]. TOUGH2 numerically simulates coupled non-isothermal heat and fluid transport through porous media based on an adaptation of Darcy's Law. Thermodynamic conditions according to the steam-table equation [International Formation Committee, 1967] are used to calculate phase transitions, so that gases and liquids can be included. Different equations of state (EOS) simulate different components: here we used EOS3 (water, air) to simulate the vadose zone and EOS1 (water) to simulate the saturated zone. Full details of the model can be found in Pruess et al. .
5.3. Vadose Zone
 Our various geophysical data and models provide constraints on the shallow hydrothermal system that can be used in a numerical model. Transient electromagnetic soundings throughout the caldera indicate that the groundwater table is approximately 250 m beneath the surface in the study area [MacNeil et al., 2007], although there is also some evidence of shallower groundwater [Mauri et al., 2012]. Magnetic data and models are consistent with a faulted terrain, dominated in this area by steeply dipping (60°) normal faults. SP and CO2-flux data indicate that these faults serve as barriers to flow, rather than enhancing flow along the fault plane, and that CO2 flux anomalies average around 2 × 10−5 kg/m2s (Table 1). Shallow temperature measurements recorded over a period of several years at five different depths (0.15 to 1.5 m) at Comalito cinder cone show average temperatures of between 46 and 74°C, and maximum temperatures of between 61 and 82°C (Figure 6) [see Pearson et al., 2008]. Thus the hydrothermal system should be locally dominated by upward flow of fluids through the footwalls of steeply dipping faults, with near-surface temperatures of ∼70°C.
 We created a rectangular, 2-dimensional grid 1000 m wide and 250 m deep (Figure 7c). In the x-direction the grid had 74 elements; 60 cells of 5 m in the center to represent the 100-m wide fracture zone and 100 m on either side, and 7 elements of 50 m adjacent to each vertical boundary (Figure 7). In this way, the vertical boundary conditions should not affect the area of interest within the fracture zone. The vadose zone above the approximately 250-m-deep water table [MacNeil et al., 2007] was represented by 25 layers of 10 m. Within the grid, less-permeable faults dipped at approximately 60° (Figure 7c).
 Rock properties were measured by Chiodini et al. , MacNeil et al. , or are typical of fractured basalt (Table 2). Bulk permeability within the fracture zone was determined by comparison of model outputs with observed surface temperatures (see Tables S1 and S2 in the auxiliary material). Permeability of the fault was set at the same value as the wall rock for simplicity, as both are relatively impermeable compared to the fracture-zone rock. The model was run for 2500 years, approximately the length of time since the caldera formed [Williams, 1983].
The permeability of the fracture-zone rock was varied as described in the text and theauxiliary material.
Permeability of wall rock
1 × 10−16
Permeability of fault rock
1 × 10−16
Wet heat conductivity
 The model was initially at atmospheric pressure and temperature (101 kPa and 25°C) with 80% air and 20% water. The upper boundary of the model domain had the same pressure and temperature conditions, but with 99% porosity and a very large volume to represent the atmosphere. This effectively fixed the temperature and pressure at atmospheric values. Both of the vertical boundaries were impermeable to flow. The lower boundary condition was set to represent the water table, and so was 100% water.
 The lower boundary condition is the major constraint of the model as it dictates how, at what temperature, and at what rate, fluids flow through the fracture zone. As the fracture zone appears to host all heat and fluid flow, this was the only area where different boundary conditions were used. Initially the pressure was atmospheric, and then water and air were injected at a temperature and rate determined by comparison of model outputs with observed surface temperatures. Either enthalpy or temperature could be prescribed; by making the rock density (and therefore specific heat capacity) arbitrarily large (2 × 1040 kg/m3) we fixed the injection temperature, thereby avoiding problems with enthalpy during phase changes. Temperatures of 70 and 120°C were used to ensure that injection of both water and water vapor were modeled (see Tables S1 and S2). Air and water were injected in equal parts as suggested by gas geochemistry [Chiodini et al., 2005]. Air was used to represent the gases in the vadose zone for a number of reasons: the TOUGH2 model can only work with water and either air or CO2, gas geochemistry suggests that at Masaya the gases are ∼45% air, 45% water, and 10% CO2 [Chiodini et al., 2005]; and we have found that the air module is more computationally stable. As the thermal properties of air and CO2 are quite similar compared to water or water vapor, we believe that the differences between the two models would not be significant.
 A range of simulations were run with fracture-rock permeability of between 1 × 10−14 and 1 × 10−10 m2and air-water injection rates of between 1 × 10−7 and 1 × 10−3 kg/m2s (Figure 8). After 2500 years of fluid injection, the air, water and water vapor mixture rise steadily, primarily controlled by the geometry of faults (Figure 7d). Most fluid gets redirected shallowly to travel along the footwall of the west fault, but some also travels between the faults (Figures 7b and 7d). Temperatures correlate with gas flux, increasing near the footwall of the west fault. Simulated fluid flux is in good agreement with fluid flux inferred from SP (Figures 7a and 7b).
 Permeability is unknown, but can be partially constrained by the known near-surface temperature. Modeling shows that a range of conditions can approximate the temperature, but that some heat is lost in transport and the base temperature must therefore be above 70°C (Figure 8b). A permeability range of between 1 × 10−11 and 1 × 10−14 m2is possible for near-surface maximum temperatures of 70°C (Figure 8a), although the low injection rate of between 1 × 10−7 and 1 × 10−6 kg/m2s required with a permeability of 1 × 10−11 m2 is not supported by surface gas flux measurements. The higher injection rates of >1 × 10−5 kg/m2s that result in 70°C maximum near-surface temperature with other permeabilities are however in good agreement. For a low permeability of 1 × 10−13 m2 or less, fluid flow (particularly water) is inhibited and the system only heats up with a high injection rate (Figure 8a). In the extreme case of a permeability of 1 × 10−14 m2 and gas flux of 1 × 10−3 kg/m2s, the fluid is trapped and pressure builds up beyond physically realistic values so that the model crashes (Figure 8b). The unusually high gas fluxes observed at Masaya volcano [Lewicki et al., 2003] suggest that fluid can flow freely through this fracture zone, and therefore we believe that a permeability of 1 × 10−12 m2 is the most appropriate. This is typical of basalt, which is in the range of 10−14 to 10−9 m2 [Freeze and Cherry, 1979; Saar and Manga, 1999; Ingebritsen and Manning, 2010].
 The injection rate at depth also needs to be determined from modeling, and comparisons with surface flux values allow us to refine the likely range. Field measurements suggest a surface CO2 flux of 2 × 10−5 kg/m2s, and as 10% of the measured gases are CO2 [Chiodini et al., 2005], this would imply a total fluid flux of 2 × 10−4 kg/m2s. Given a permeability of 1 × 10−12 m2, the models suggest that an injection rate of ∼1 × 10−4 kg/m2s would give rise to 70°C surface temperatures (Figure 8a), in good agreement with these surface measurements.
 Using these models we can estimate a number of surface parameters, and compare them with measured values where available. Using an injection rate of 2 × 10−4 kg/m2s, we find that the optimum temperature at the water table is 96°C (Figure 9a), which would allow some boiling to occur creating the water vapor that is visible at the surface. The maximum gas flux in the layer below the atmosphere is 1.7 × 10−4 kg/m2s (Figure 9b), close to the 2 × 10−4 kg/m2s measured but suggesting that the injection rate at depth may be higher than 2 × 10−4 kg/m2s. The maximum modeled surface heat flux is 162 W/m2 (Figure 9c), higher than the 91 W/m2 measured by Chiodini et al. , but feasible given the uncertainty associated with heat flux measurements.
 Fluid velocity is also affected by permeability and heat input, and these models allow us to estimate travel times. Maximum gas velocities vary from 5 × 10−5 m/s in the low-permeability models to 5 × 10−3 m/s in a system with a high permeability of 10−10 m2 and a fluid input of 1 × 10−3 kg/m2s. This implies travel times ranging from 60 days to 14 hours between the water table (250 m depth) and the surface. The velocity for our optimum model (permeability 1 × 10−12 m2, fluid injection 2 × 10−4 kg/m2s) is 6.7 × 10−4 m/s, resulting in a minimum travel time of ∼4 days for fluids to travel through the system.
 For these models, the permeability of the wall and fault rock have been assumed to be relatively impermeable at 1 × 10−16 m2, but this is an arbitrary value. For a permeability of 1 × 10−14 m2 or less the models appear to be relatively insensitive to permeability (Figure 10). Unfractured basalt is typically on the order of 1 × 10−14 m2 or less [Freeze and Cherry, 1979]. For fault permeability close to the permeability of the fracture, the near-surface temperature of 70°C is not reached (Figure 10b) because circulation increases and the fluid flow distribution is not focused, supporting our theory that relatively impermeable faults are creating barriers to fluid flow.
 The magnetic models suggest a shallow scoria layer underlain by lava. However, a uniform permeability distribution appears to give the best fit to the data. Adding a more permeable shallow layer results in cooler temperatures near the fault, while a less permeable layer results in hotter temperatures (Figure 10). We infer that the variation in permeability between fractured lava and scoria is less significant than their very different magnetic properties.
5.4. Saturated Zone
 The degassing system at Masaya volcano is unusually stable and open, and the distribution of diffuse degassing can be used to infer variations in fluid flux at depth. The three distinct zones of diffuse degassing along the fracture zone on the flank of Masaya volcano (Figure 1) are all characterized by stable and consistently elevated temperature and CO2 flux. Soil temperatures at the area of diffuse degassing at Comalito cinder cone range between 55 and 80°C [Pearson et al., 2008], and are similar to those measured at the degassing area closest to the crater (38 to 83°C). Measured CO2 fluxes are similar at all three degassing zones, varying between 2.3 and 2.6 × 10−5 kg/m2s. At Comalito cinder cone, we measured gas fluxes that were stable over remarkably long periods of time, varying by less than 10% from average over 5 months. The uniformity in temperature and CO2 flux across the three degassing zones suggests that they share a common source. Gases at Comalito cinder cone, 3.5 km from the active crater, retain a magmatic component [Lewicki et al., 2003], but as deep groundwater flow is from Comalito toward the active crater [MacNeil et al., 2007] there must either be separate magmatic sources, or one extensive source at depth. Lava ponding in the active crater in 2006 corresponded to changes in temperature at the Comalito cinder cone diffuse degassing zone [Pearson et al., 2008], suggesting that there may in fact be an aerially extensive magmatic source at depth beneath the caldera.
 For three distinct degassing zones to occur with one magmatic source, the source must either be focused by variations in permeability or from convection of groundwater. To determine whether the consistent spacing, temperatures and gas fluxes in the distinct degassing zones can be explained by groundwater convection, we created a TOUGH2 model of the saturated zone. We used a grid of 35 by 110 cells extended over 3500 m depth and 5500 m length respectively (Figure 11a). The model was 2D; the third dimension was one cell of 100 m to represent the thickness of the fracture zone. Horizontally the fracture was modeled as the 3500 m-long high-permeability zone, plus 1000 m of lower permeability on either side (Table 2 and Figure 11a) to ensure that the boundary conditions did not affect the model. A depth of 3500 m was assumed for simplicity, as there is no information to constrain depth to the source. The entire model was fully saturated at atmospheric temperature and pressure initially. The top boundary was fixed at atmospheric conditions, and the vertical boundaries were no flow. Heat or fluid was injected at various rates along the bottom boundary to replicate the volcanic source. This model was also run for 2500 years.
 TOUGH2 modeling shows that a range of conditions can give rise to convection along the length of the fracture system (Figure 11 and Table S3). A permeability of between 1 × 10−14 and 1 × 10−11 m2 was modeled with either heat or fluid injected at a variety of rates (Figure 12). Heat was injected at between 0.1 and 50 W/m2. Fluid was injected at an enthalpy of 632 or 1300 kJ/kg (∼150°C or 300°C respectively), and with an injection rate of between 1 × 10−6 and 1 × 10−4 kg/m2s. With a permeability of 10−14 m2 the system was generally conductive, but for higher permeability convection readily occurred.
 The water table temperature estimated in the vadose-zone model provides a constraint when determining the permeability and injection rate within the saturated-zone system. When injecting heat, the permeability must be between 1 × 10−13 m2 and 1 × 10−12 m2for the near-surface temperature of 96°C to be reached (Figure 11a). The heat injection rate must be around 5 W/m2 for the lower permeability, and greater than 20 W/m2 for the higher permeability (Figure 12a). When fluid is injected, 96°C can be reached with any permeability, but with an injection rate of 1 × 10−5 kg/m2s or higher for 300°C fluid (Figure 12b), and closer to 1 × 10−4 kg/m2s for 150°C fluid (Figure 12c). Therefore any of these scenarios are theoretically possible.
 Three distinct diffuse degassing zones are observed at Masaya volcano (Figure 1), and therefore the model needs to test if three distinct plumes can be produced. This can be done with a range of heat injection, although the number of plumes does not vary directly with injection rate (Figure 13a). The higher the heat injection rate, the larger the number of plumes for a permeability of 1 × 10−11 m2 but it is extremely unlikely that the saturated zone is more permeable than the vadose zone. There are only two heat injection rates that result in three plumes for a permeability of 1 × 10−12 m2; 1 W/m2 and 20 W/m2 (Figure 13a). Both of these result in surface temperatures that are too cold (Figure 12a). For a permeability of 1 × 10−13 m2, injection rates of 5 W/m2 or higher result in plumes that reach the surface, and in some cases with 3 plumes (Figure 13a). An injection rate of 5 W/m2 gives a groundwater table fluid temperature of ∼90°C (Figure 12a) and therefore this could be a possible source of the groundwater convection. This results in three regularly spaced, consistent convective plumes (Figure 11b).
 It appears that hot water entering the groundwater system at 3500 m depth also results in groundwater convection with three plumes. For a hotter fluid at around 300°C, it requires a fairly high permeability of 1 × 10−4 kg/m2s for three plumes to develop (Figure 13b). However, with a permeability of 1 × 10−12 m2 and injection of ∼150°C fluid, three plumes develop irrespective of fluid injection rate (Figure 13c). A groundwater table fluid temperature of 96°C results from a fluid injection rate of 5 × 10−5 kg/m2s (Figure 12c). This results in a more random pattern of convection than with simple heat injection (Figure 11c), but it is stable over 1000 years.
 Although modeling shows that both a uniform heat source and a uniform fluid source at depth can produce three zones of diffuse degassing, comparison of heat and fluid flow rates allow us to determine which is more likely. When heat is injected at 5 W/m2 into rock with a permeability of 1 × 10−13 m2, the maximum heat flux in the layer below the upper boundary layer is 17 W/m2 and the maximum fluid flux is 5 × 10−5 kg/m2s; significantly lower than the vadose zone model would suggest. However, when water is injected at 5 × 10−5 kg/m2s with an enthalpy of 632 kJ/kg into rock with a permeability of 1 × 10−12 m2, the maximum heat flux is 195 W/m2 and the maximum fluid flux is 5 × 10−4 kg/m2s. These are in good agreement with vadose zone values of 2 × 10−4 kg/m2s and heat flux of 162 W/m2. Therefore it seems entirely possible that hot fluid entering the groundwater system at around 3500 m is driving groundwater convection that results in the three distinct zones of diffuse degassing observed at Masaya volcano.
 Interpretation of magnetic traverses on the flank of Masaya volcano suggest that SE-dipping normal faults redirect upward fluid movement (Figure 4). Positive SP and CO2 anomalies to the NW, in the footwall of the main fault, suggest that these are sealing faults, as does a complete absence of CO2 and a small SP anomaly to the SE (Figures 3–5). Numerical modeling shows that this has a strong effect on mass flux, soil gas concentrations, and soil temperature (Figure 7b and 7d). The small negative SP anomaly to the SE could result from fluid flow between the faults, as seen in the model, or from near-surface permeable sediments and/or fractured bedrock channeling fluid away from the fault. However, the large positive SP anomaly observed in the footwall of the west fault can only be explained by the sealing faults [Goff and Janik, 2000; Hochstein and Browne, 2000].
 Normal faults dipping at 60° and extending to some depth can explain the distinct distribution of CO2 and SP anomalies observed at the surface (Figure 7). More steeply dipping faults would result in sharper and more localized anomalies [Goff and Janik, 2000; Hochstein and Browne, 2000]. For a water table at 250 m depth, a 60° normal fault would channel fluids to the surface from an approximately 150 m wide area. This is roughly consistent with the dimensions of the top of a single convection cell in our saturated-zone modeling (Figure 11).
 Numerical models are influenced by system geometry, heat injection rate, and rock permeability. Although the fracture geometry at Masaya volcano is constrained by surface features and magnetic surveys, heat injection rate and permeability are unknown. Permeability of basalt is typically in the range of 10−14 to 10−9 m2 [Freeze and Cherry, 1979; Saar and Manga, 1999; Ingebritsen and Manning, 2010]. Our inferred permeability of 1 × 10−12 m2 in both the vadose and saturated zones is within the range of measured basalt permeabilities worldwide. It is also within the range of 7.7 × 10−10 to 6.4 × 10−15 m2 estimated by MacNeil et al.  for groundwater models of Masaya volcano. Generally however, the permeability of basalt decreases with depth [Ingebritsen and Scholl, 1993]. With the heat injection model, a lower permeability resulted in three plumes (Figure 13a), suggesting that both heat conduction and convection may be important at depth. In either case, the fracture zone is still relatively permeable, possibly explaining why it hosts the only extensive hydrothermal feature with surface expression outside of Santiago crater at Masaya volcano [MacNeil et al., 2007]. Such focused degassing has also been observed at other volcanoes [Chiodini et al., 2001].
 Although faults are often assumed to be permeable conduits to flow [e.g., Evans et al., 2001; Finizola et al., 2002; Barde-Cabusson et al., 2009], the faults in our vadose zone models are relatively impermeable. Previous studies have shown that permeability can vary by several orders of magnitude within a fault [Manzocchi et al., 2008], and/or can create a predominantly low-permeability zone with channels of high permeability [Marler and Ge, 2003; Fairley et al., 2003]. Permeability is altered by fracture dilatancy, cementation, compaction, clay formation, deformation stress and temperature histories, micro-cracking of dense rock, chemical alteration, diagenesis, and mineral precipitation [Goeff and Gardner, 1994; Goff and Janik, 2000; Fisher and Knipe, 2001; Wibberley and Shimamoto, 2003; Bernabé et al., 2003]. Low-permeability fault gouge may be present, as at Elkhorn Fault Zone, South Park Colorado, USA [Marler and Ge, 2003]. Fault geometry can also be important. Caine et al. showed that a fault zone can be divided into fault core, fractured damage zone, and undeformed protolith. The fault core may be clay-rich, brecciated and/or geochemically altered, resulting in a low permeability. If the fault core is large relative to the fault as a whole, the overall permeability will be low. In addition, Caine et al. showed that small damage width of a fault (relative to its total width) will result in localized strain which inhibits flow, acting as a low-permeability zone. There are therefore a number of ways that the faults at Masaya volcano could behave as low-permeability zones that redirect fluid flow.
 To recreate the surface temperatures observed at Masaya volcano we had to estimate both permeability and injection rate. When an equal mixture of water and air are injected into the base of a vadose-zone model with permeability of 1 × 10−12 m2 at 2 × 10−4 kg/m2s, maximum surface heat flux is 162 W/m2 (Figure 9c), within an order of magnitude of the measured heat output of 91 W/m2 at Comalito cinder cone [Chiodini et al., 2005]. The average heat flux over the fault is 74 W/m2, in excellent agreement with previous measurements given the uncertainty in heat flux measurements and the difficulty in locating the maximum heat flux in the field. This is also representative of volcanoes worldwide [Harris and Stevenson, 1997].
 Numerical modeling shows that even with totally uniform fluid injection, three distinct and stable degassing zones can develop along a fracture zone (Figure 11). This requires a fluid injection rate of 5 × 10−5 kg/m2s, resulting in a water table fluid flux of 5 × 10−4 kg/m2s. This is in good agreement with the 2 × 10−4 kg/m2s implied from the 10% gas content of CO2 measured in the three distinct degassing zones, but less than previous CO2 measurements of 5.8 × 10−4 kg/m2s (implying 5.8 × 10−3 kg/m2s total gas) at Comalito cinder cone would suggest [Lewicki et al., 2003]. Although a heat injection rate of 5 W/m2is also theoretically possible to explain the three distinct degassing zones, this results in heat and gas fluxes an order of magnitude smaller again and therefore seems less likely. Total modeled heat output of 11 MW along the 100-m wide fracture zone is similar to estimates for thermal zones at Karapiti geothermal field, New Zealand [Hochstein and Bromley, 2001] and Vesuvius, Italy [Chiodini et al., 2005]. The total heat output for one degassing zone is between 2.8 and 6.3 MW; higher than the 0.9 MW estimated at Comalito but similar to values at Vulcano and Pantelleria, Italy [Chiodini et al., 2005].
 We believe that our models adequately explain the observed features at Masaya volcano, despite a number of simplifications. As mentioned previously, air is used to represent both air and CO2, although we believe that this is justified by their relatively similar thermodynamic properties. However, topography, variations in hydraulic properties, and lateral heterogeneities along and within the fracture may influence fluid flow in unidentified ways. For example, magnetic and SP measurements suggest that the width of the fracture zone varies along its length. Work by Méheust and Schmittbuhl  and Neuville et al.  has shown that fracture roughness can enhance or inhibit flow. We think that these complex heterogeneities play a role in channeling groundwater and gas flow, but that convection is also a fundamental control.
 Geophysical measurements and numerical modeling show that fluid flux at Masaya volcano is controlled by faults on a local scale at very shallow levels (100s of meters), and by groundwater convection in a flank fracture on the kilometer scale (Figure 14). Magnetic measurements in 2007 detected a NE-SW trending anomaly of 2300 nT to the east of our study area that coincides with surface topographic offset. Modeling of this data suggests relatively impermeable faults dipping SE at 60°. Elevated SP and CO2 flux to the NW of the fault, and small SP values and an absence of CO2 to the SE suggest that these faults are barriers to flow. Numerical models show that with a permeability of 1 × 10−12 m2in the fracture-zone rock, water table temperature of 96°C and fault rock permeability of less than 1 × 10−14 m2, surface gas flux and temperature measurements can be matched. The models show that fluid flux increases toward the footwalls of the faults and becomes negligible over them, as suggested by the CO2 and SP measurements. Gases in the model take a minimum of 4 days to travel from the groundwater table to the surface.
 Numerical models show that the three diffuse degassing zones observed within the 3–4 km fracture on the flank of Masaya volcano can be explained by convection of groundwater within the saturated zone. The number of convection cells and their dimensions are dependent primarily on rock permeability and heat or fluid injection rate for a given geometry. It was found that a permeability of 1 × 10−12 m2 and a uniform water injection rate of 5 × 10−5 kg/m2s at an enthalpy of 632 kJ/kg could best explain the gas flux and temperature suggested by the vadose zone model, and the three distinct degassing zones. This model results in a total heat output of 11 MW along the fracture zone, and between ∼3 and 6 MW at an individual degassing zone.
 From our geophysical surveys and models we conclude that within a generally high-permeability fracture zone on the flank of Masaya cone there are faults and other variations that create localized low-permeability zones. At the central degassing zone, heat and fluid rising from depth due to groundwater convection are redirected from a ∼150-m-wide area because of faults. Relatively uniform heat and fluid transport from depth is focused by groundwater convection and permeability variations at several scales in the upper >3 km of the caldera section. A good understanding of the subsurface geology and hydrologic system is therefore vital when attempting to understand the sources of both spatial and temporal variations in diffuse degassing.
 We would like to thank staff at Instituto Nicaragüense de Estudios Territoriales. Field assistance from S. Kruse, P. Wetmore, L. Connor and students in the University of South Florida Volcanology field class is gratefully acknowledged, as is help from students at the Centro de Investigaciones Geocientificas at the National Autonomous University of Nicaragua. This manuscript was improved by reviews from Steven Ingebritsen, William Evans, Stéphanie Barde-Cabusson, Micol Todesco and Glyn Williams-Jones. The work was partially funded through the USGS water resources program.