Differences in the thermal state of subducting crust along the trench of a subduction zone cause differences in subduction zone temperature that persist to tens of kilometers down-dip of the trench. The resulting differences in fluid viscosity, permeability, and hydraulic conductivity can lead to trench-parallel variations in fluid pressure on the plate boundary fault. Temperature differences in locations with low décollement temperature (<75°C) at the trench result in large differences in fluid viscosity and fluid flow from the cold to the hot side of the system. Margins characterized by high décollement temperature at the trench are probably dominated by temperature-controlled differences in intrinsic permeability, resulting in fluid flow from the hot to the cold side of the system. Margins with large trench-parallel temperature differences support considerable trench-parallel fluid flow, and the effect is accentuated for cases where compaction driven dewatering is concentrated near the trench. Type locations for along-strike hydrologic differences include margins with subducting crust with differences in plate age along the trench or patchy hydrothermal circulation. In such cases, the three-dimensional pattern of subduction zone fluid flow should be considered when inferring the location of the source of fluids sampled from boreholes or seeps. Additionally, trench-parallel differences in fluid pressure control effective stress on the plate boundary fault. Reduced effective stress on the plate boundary on the low hydraulic conductivity side of a subduction zone reduces the effective friction coefficient, which both reduces frictional heating and may delay the onset of frictional instability.
The distribution of fluid pressure within subduction zones controls spatial variations in both effective stress and fluid flow velocity. Through its control on effective stress, fluid pressure affects fault strength [Byerlee, 1990; Davis et al., 1983; Hubbert and Rubbey, 1959] and possibly the position of the updip (seaward) limit of the seismogenic zone [Moore and Saffer, 2001; Scholz, 1998]. The updip limit of subduction zone seismicity and coseismic slip affect tsunami generation [Satake and Tanioka, 1999] and the rupture area of earthquakes [Hyndman et al., 1997]. On the basis of the coincidence of the updip limit and temperatures of ∼150°C estimated from thermal models of several subduction zones, thermal state is generally thought to drive diagenetic and metamorphic reactions that trigger the transition to seismogenic faulting [e.g., Hyndman et al., 1997; Moore and Saffer, 2001]. Within subduction zones, large spatial variations in fluid pressure and effective stress [Bangs et al., 2004; Bekins et al., 1995; Moore et al., 1998; Saffer and Bekins, 1998; Screaton et al., 1990; Spinelli et al., 2006; Stauffer and Bekins, 2001] are controlled by the permeability distribution and the magnitude and distribution of fluid sources [Screaton et al., 1990; Saffer and Bekins, 2002, 2006]. Most previous modeling studies designed to gain insight into fluid pressures and fluid flow in subduction zones have focused on vertical cross sections perpendicular to the trench (Figure 1) [Bekins et al., 1995; Henry, 2000; Saffer and Bekins, 1998; Screaton et al., 1990; Spinelli et al., 2006]. These studies, along with field observations [e.g., Saffer, 2003; Screaton et al., 2002], indicate that excess fluid pressure (fluid pressure above hydrostatic) along the plate boundary fault (décollement) generally increases with distance landward of the trench. Although trench-perpendicular cross sections capture the large down-dip variations in fluid pressure, they do not incorporate potentially significant trench-parallel variations (Figure 1). For example, an abrupt trench-parallel change in the geometry of the Barbados Ridge accretionary complex near the Tiburon Rise likely results in trench-parallel differences in fluid pressure that drive fluid flow oblique to the trench [Cutillo et al., 2003; Screaton and Ge, 1997]. Temperature can also have important effects on fluid pressure through its effects on permeability, fluid density, and fluid viscosity. The causes of local temperature differences within subducting crust (e.g., hydrothermal circulation, off-axis volcanism, and differences in plate age) are common globally. Any resulting trench-parallel differences in fluid pressure should control the origin of fluids sampled near the trench by driving trench-parallel fluid flow and also play a role in altering fault strength. Geochemical investigations of décollement fluids can constrain temperature windows for key reactions [e.g., Chan and Kastner, 2000; Kimura et al., 1997] that are commonly used to estimate a position within a subduction zone for the fluid source [e.g., Silver et al., 2000]. In settings with the potential for a large trench-parallel component of fluid flow (particularly across thermal transitions) the three-dimensional fluid flow field will influence inferences of fluid source location.
In this paper, we focus on spatial variations in fluid pressure caused by trench-parallel differences in temperature. We show that these temperature differences alone can cause differences in fluid pressure and affect flow patterns. We use three-dimensional numerical models of fluid flow to examine the controlling parameters of trench-parallel fluid flow. We estimate the potential importance of trench-parallel flow for a variety of subduction zone settings and discuss these results in the context of constraints from active margins with trench-parallel thermal variability.
2. Subduction Zone Temperature and Fluid Properties
The ease with which fluid flows through a material (i.e., hydraulic conductivity; K) is a function of the intrinsic permeability of a material (k), fluid density (ρ), and fluid viscosity (μ):
where g is acceleration due to gravity. Over the range of estimated temperatures (2–400°C) and fluid pressures (40–240 MPa) in the shallow portion of most subduction zones, fluid viscosity varies from 0.0016 to 0.00011 Pa·s and fluid density varies from 1020 to 819 kg/m3 [Parry et al., 2000]. The large reduction in fluid viscosity with increasing temperature (>90% reduction from 2 to 400°C) has a much larger influence on hydraulic conductivity than the change in fluid density (20% reduction) over the same temperature range. Raising the fluid temperature in most of this pressure and temperature range results in increased hydraulic conductivity, for a given permeability.
The intrinsic permeability of a material may also be affected by variations in temperature. Experimental results demonstrate that permeability is reduced with increased temperature due to healing of microfractures and formation of authigenic minerals in pore spaces [Kato et al., 2004; Morrow et al., 2001; Tenthorey and Fitz Gerald, 2006; Yasuhara et al., 2006]. These effects are most pronounced at low porosity [Germanovich et al., 2001; Lowell et al., 2003]. Therefore, in typical subduction zone material, temperature-dependent permeability is likely most prominent at depths greater than ∼8 km below the seafloor [Bray and Karig, 1985]. In their examination of subduction zone fault rocks, Kato et al.  found a decrease in permeability of approximately one order of magnitude upon increasing temperature from 30°C to 250°C. The trend of this temperature controlled permeability reduction is similar to a trend found for fractured granite [Morrow et al., 2001].
Subduction geometry, convergence rate, frictional heating, and the thermal state of the subducting oceanic crust have been recognized as primary controls on subduction zone temperature at the regional scale [e.g., Molnar and England, 1995], whereas forearc thermal conductivity, radiogenic heating, and fluid flow play smaller roles [Dumitru, 1991; Peacock, 1987; Wang et al., 1993]. Wang et al.  found that sediment thickening within an accretionary wedge has a greater influence on temperatures than diffuse fluid expulsion from the wedge. In the underthrust section, the volume of fluid produced by sediment compaction and metamorphic reactions is not sufficient to pervasively alter subduction zone temperatures [e.g., Peacock, 1987; Reck, 1987]. Temperatures along isolated preferential fluid flow paths may be altered from background, conductive values, but the overall thermal effect of fluid flow either up through the margin wedge or out of the system along the décollement is minor [Peacock, 1987].
In this study, we focus primarily on the potential for trench-parallel differences in subduction zone temperature due to differences in the thermal state of the incoming crust. Therefore we isolate the effect of trench-parallel temperature differences from the effect of variations in material dewatering due to differences in subduction geometry and convergence rate [e.g., Bekins and Dreiss, 1992; Saffer and Bekins, 2006] on the three-dimensional pattern of fluid flow. We examine the effects of temperature-dependent variations in both permeability and fluid properties. Reduced fluid viscosity on the hot side of a system will increase hydraulic conductivity. However, the reduced permeability on the hot side will tend to decrease hydraulic conductivity relative to the cold side of a system. Where these opposing factors do not precisely counterbalance each other, some fluid (being driven out of the subduction zone system by sediment compaction and dewatering) should flow across the thermal transition. Where variations in fluid properties dominate, fluid will flow from the cold side of the system to the hot side; where variations in permeability dominate, fluid will flow from the hot side of the system to the cold side.
3. Causes of Trench-Parallel Temperature Differences
Causes of trench-parallel temperature differences of subducting crust include hydrothermal circulation, off-axis volcanism, and differences in plate age. The thermal state of ocean crust is typically related to crustal age; heat flux from conductively cooled lithosphere is estimated as
where q is heat flux (mW/m2) and t is crustal age (Ma) [Harris and Chapman, 2004, and references therein]. Fracture zones offsetting ocean crust of different ages can result in large trench-parallel differences in thermal state of crust approaching and entering a subduction zone, particularly in locations with young, rapidly cooling crust. Additionally, spatial variations in volcanism or hydrothermal circulation can lead to trench-parallel differences in the thermal state of subducting crust. Local volcanism can add heat to ocean crust. Hydrothermal circulation in fractured basaltic basement extracts and redistributes crustal heat. We examine the potential for trench-parallel differences in subduction zone temperature due to differences in plate age (Chile, Mexico, and Cascadia), local volcanism (Nankai), and hydrothermal circulation (Costa Rica; Figure 2).
3.1. Differences in Plate Age: Chile, Mexico, and Cascadia
At the Peru-Chile Trench (Figure 3), the Guafo Fracture Zone separates ∼6.5 Ma crust to the south from ∼15 Ma crust to the north [Tebbens et al., 1997]. Although seafloor heat flux observations in the area are sparse [Cande et al., 1987], estimates of the thermal state of the margin wedge from the depth of a gas hydrate bottom simulating reflector show a trend of higher heat flux to the south “related to the younger age of the incoming oceanic lithosphere” [Grevemeyer et al., 2003]. On the basis of a globally averaged conductive lithospheric cooling model [e.g., Harris and Chapman, 2004], heat flux from the 6.5 Ma crust is expected to be ∼200 mW/m2; heat flux from the 15 Ma crust is expected to be ∼130 mW/m2.
The subduction zone at the Peru-Chile Trench separates the subducting Nazca Plate from the overlying South American Plate. Where the Guafo Fracture Zone enters the Peru-Chile Trench, the convergence rate is ∼8.4 cm/yr [DeMets et al., 1990]. The wedge surface slope is ∼6.4° and the décollement dips ∼2.3° [Behrmann and Kopf, 2001]. Approximately 700 m of the 3400 m of incoming sediment is scraped off into the accretionary prism [Behrmann and Kopf, 2001].
At the Middle America Trench, the Rivera Fracture Zone and a fracture at ∼18°N offset the East Pacific Rise (Figure 4). In a zone of distributed shear between the fractures, magnetic anomalies (and therefore plate ages) are not easily identified [Klitgord and Mammerickx, 1982]. For this study we compare subduction zone temperatures north of the Rivera Fracture Zone (9.5 Ma crust) and south of the 18° offset (4 Ma crust) [Klitgord and Mammerickx, 1982; Lonsdale, 1995]. Seafloor heat flux observations in the area are sparse, but the one measurement in the study area on the incoming crust south of the 18° offset (213 mW/m2 [Vacquier et al., 1967]) is fairly consistent with a global lithospheric cooling model for 4 Ma crust (i.e., 250 mW/m2). On the basis of the same model, heat flux from the 9.5 Ma crust north of the Rivera Fracture Zone is expected to be ∼160 mW/m2.
The subduction zone separates the subducting Cocos and Rivera Plates from the overlying North American Plate. Along the trench, the convergence rate increases from 1.9 cm/yr at ∼20°N, to 2.9 cm/yr at 19°N, and 3.8 cm/yr at 18.2°N [DeMets and Wilson, 1997]. The wedge surface slope is ∼4° and the décollement dips ∼13° [Pardo and Suarez, 1993; Moore et al., 1982]. Approximately 600 m of the 825 m of incoming sediment (much of the sandy trench turbidites) is incorporated into the accretionary prism [Moore et al., 1982].
3.1.3. Cascadia (Blanco Transform)
The Blanco Transform offsets the Juan de Fuca Ridge from the Gorda Ridge by >300 km (Figure 5). At the Cascadia trench, the Blanco Transform separates 9.5 Ma crust to the north from 5 Ma crust to the south [Wilson, 1993]. Immediately north of the transform at the trench, magnetic anomalies indicate a sliver of >10 Ma crust with magnetic lineations oblique to the trench [Wilson, 1993]. For this study we compare subduction zone temperatures immediately south of the transform and ∼160 km north of the transform, where magnetic lineations parallel the trench. Heat flux from the 5 Ma crust is expected to be ∼225 mW/m2; heat flux from the 9.5 Ma crust is expected to be ∼160 mW/m2, on the basis of the lithospheric cooling model. Both south and north of the transform, seafloor heat flux observations are generally consistent with the model estimates, ∼240 mW/m2 to the south [Korgen et al., 1971] and ∼150 mW/m2 to the north [Korgen et al., 1971; Moran and Lister, 1987].
The subduction zone separates the Juan de Fuca Plate from the overlying North American Plate. At the Blanco Transform, the convergence rate is ∼3 cm/yr [DeMets and Dixon, 1999]. The wedge surface slope is ∼2.8° and the décollement dips ∼1.5° [Gerdom et al., 2000; Gulick et al., 1998]. A thick sediment section approaches the trench on the incoming plate. South of the transform, approximately 1300 m of the 1540 m of incoming sediment is incorporated into the accretionary prism [Gulick et al., 1998]. North of the transform, approximately 1600 m of the 3500 m of incoming sediment is scraped off into the margin wedge [MacKay, 1995; MacKay et al., 1992; Trehu et al., 1995].
3.1.4. Cascadia (Vancouver Island)
At the Cascadia trench offshore Vancouver Island and Olympic Peninsula (Figure 5), the convergence direction [DeMets and Dixon, 1999] is oblique to magnetic lineations formed at the Juan de Fuca Ridge [Wilson, 1993]. As a result the age of the subducting crust at the trench increases to the southeast, from ∼6.5 Ma off Vancouver Island to ∼9.5 Ma offshore Olympic Peninsula [Wilson, 1993]. The younger and warmer crust entering the trench off Vancouver Island should lead to a warmer subduction zone there than offshore Olympic Peninsula. For the warm and cool portions of the northern Cascadia subduction zone, we use previously published modeled temperatures for the décollement for Vancouver Island and Olympic Peninsula [Oleskevich et al., 1999]. We note the convergence rate and subduction zone geometry for comparison to other margins and underthrust fluid source distributions. Off Vancouver Island, the Juan de Fuca Plate subducts beneath the North American Plate at ∼3.6 cm/yr [DeMets and Dixon, 1999]. The slope of the seafloor on the wedge is 2.8°; near the trench, the subducting plate dips 3.5° [Yuan et al., 1994]. The incoming sediment thickness is ∼3500 m off Vancouver Island and ∼3200 m off Olympic Peninsula, with nearly all the sediment assumed to be incorporated into the margin wedge [Oleskevich et al., 1999].
3.2. Off-Axis Volcanism: Nankai
The thermal history of the subducting plate at the Nankai Trough (Figure 6) is complicated and spatially variable. Seafloor spreading along the Shikoku Basin back-arc rift system stopped ∼15 Ma [Okino et al., 1994], but post-spreading volcanism continued along the Kinan Seamount Chain to 7–10 Ma [Sato et al., 2002]. Heat flow data have been used to make inferences about the anomalous thermal history of subducting crust [Wang et al., 1995; Yamano et al., 2003]. Currently, heat flux along the fossil spreading center is >180 mW/m2 [Yamano et al., 2003], higher than the ∼130 mW/m2 predicted for 15 Ma crust. Observed seafloor heat flux decreases with distance from the fossil spreading center. In the axis of Nankai Trough, approximately 140 km from the fossil spreading center, seafloor heat flux is ∼120 mW/m2 [Yamano et al., 2003].
The Nankai Trough subduction zone separates the subducting Philippine Sea Plate from the overlying Eurasian Plate. The current convergence rate is ∼4 cm/yr [Seno et al., 1993]. On the axis of the fossil spreading center ∼750 m of the 1050 m of incoming sediment is scraped off into the accretionary prism; the surface slope of the wedge is ∼1.5° and the décollement dip is ∼2.6° [Moore et al., 2001]. Approximately 140 km southwest of the fossil spreading center ∼710 m of the 1140 m of incoming sediment is incorporated into the accretionary prism; the wedge slope is ∼3.8° and the décollement dip is ∼4.2° [Moore et al., 2001].
3.3. Hydrothermal Circulation: Costa Rica
Along the Pacific margin of Costa Rica, an abrupt (≤5 km wide) transition between warm and cool crust has been delineated by heat flow surveys (Figure 7). Southeast of the transition, seafloor heat flow is 105–120 mW/m2 [Fisher et al., 2003], consistent with conductive lithospheric cooling models for ∼20 Ma crust. Northwest of the transition, hydrothermal circulation extracts heat from the crust and reduces measured heat flow values to 20–40 mW/m2 [Fisher et al., 2003]. The extent of hydrothermal cooling is controlled by the distribution of seamounts that provide points for rapid fluid exchange between the basaltic basement and the ocean [Fisher et al., 2003; Hutnak et al., 2007].
The Cocos Plate subducts beneath the Caribbean Plate at ∼8.5 cm/yr [DeMets, 2001]. Most (∼99%) of the ∼375 m thick incoming sediment column is subducted [Saito and Goldberg, 2001]. The slope of the seafloor on the wedge is 5.4°. The subducting plate dips 6° for the first 30 km of the subduction zone, after which the dip angle increases to 13° [Christeson et al., 1999].
4.1. Hydrologic Modeling
Using a three-dimensional numerical model of fluid flow within a model domain straddling a thermal transition, we examine the hydrologic effect of trench-parallel subduction zone temperature differences over a range of thermal conditions. Because intrinsic permeability and fluid properties (density, viscosity) are potentially affected by temperature, we systematically evaluate the effect of each on resulting pore pressure and flow patterns.
We use a three-dimensional finite element groundwater flow model, SUTRA3D [Voss and Provost, 2002] to simulate steady state fluid pressures, representative of long-term time-averaged pressures that relate to the overall fluid flow patterns. We model only fluid transport (not coupled fluid and heat transport) with the model. The effect of advective heat transport due to fluid flow is excluded; in subduction zones, the thermal effects of fluid flow are typically only a few degrees C and would be highly localized [Peacock, 1987]. Predetermined trench-parallel differences in temperature are used to define differences in hydraulic conductivity across a thermal transition. We determine the hydraulic conductivity for each element in the model on the basis of temperature-dependent fluid density, fluid viscosity, and permeability. Our model domain is 400 km wide (along the trench) and centered on a thermal transition. For most simulations, the thermal transition is 5 km wide, although we examine the effect of widening the transition to 100 km. The model extends from 10 km seaward to 80 km landward of the trench. The wedge taper is 11°; simulations are run for a single generic subduction zone geometry to isolate the temperature-controlled effects. Temperatures for specific margins (e.g., Chile, Mexico, Cascadia, Nankai, and Costa Rica) are used to define trench-parallel differences in hydraulic conductivity, but those individual margin geometries are not applied in the fluid flow modeling. The seaward and top faces of the model are assigned constant hydrostatic fluid pressure. The landward, bottom, and side (trench-perpendicular) faces of the model are no flow boundaries.
In order to define fluid and rock properties that define trench-parallel differences in hydraulic conductivities in our hydrologic models, we first delineate the temperature distribution within the model domain. For six settings with conditions favorable for establishing trench-parallel differences in temperature (Figure 2), we estimate décollement temperatures in order to determine the potential for hosting significant trench-parallel fluid flow. We specify hydraulic conductivities within the hydrologic model domain on the basis of temperatures both north and south of the thermal transition. For Cascadia and Costa Rica, we use existing thermal model results [Oleskevich et al., 1999; Spinelli and Saffer, 2004; Spinelli et al., 2006]. For the subset of sites where existing thermal modeling results are not available, we calculate subduction zone temperatures using the thermal model of Ferguson  [see also Bekins et al., 1995; Saffer and Bekins, 1998]. In this approach, subduction zone temperatures in 2-D are calculated by tracking a one-dimensional column as it is buried and thickened within the wedge. This 1-D transient model includes the effects of heat advection (due to the motion of the underthrust sediment relative to the overlying wedge and to the thickening of accreted sediment) and conduction during burial of subducted sediment, strain heating within a deforming accretionary wedge, frictional heating on the plate boundary fault, and radiogenic heating [Ferguson et al., 1993].
We calculate temperatures along two cross sections for each study area, one for the hotter portion of the margin and one for the colder portion of the margin. The thermal model requires estimations of heat flow on the incoming plate, convergence rate, and subduction zone geometry (décollement dip, wedge slope, sediment thickness, and proportion of sediment subducted versus accreted). These parameters for each study area are presented above (section 3). In all cases, we use thermal conductivities for pore fluid, sediment grains, and basement rock of 0.67, 3.0, and 2.5 W/m-K, respectively. The effective conductivity of the sediment is calculated using a geometric mean mixing model for the pore fluid and sediment grains [Woodside and Messmer, 1961]. We assume that temperatures in the thermal transition vary linearly between the warm and cool sections.
4.3. Fluid and Rock Properties
For the general illustration presented here, we assign permeabilities of the décollement, margin wedge, and underthrust sediment that are consistent with results constrained by laboratory data [Gamage and Screaton, 2006; Saffer and McKiernan, 2005] and previous detailed studies of individual margins [Bekins et al., 1995; Saffer and Bekins, 1998; Screaton et al., 1990]. These intrinsic permeability values are then modified to account for temperature effects; permeability decreases with increasing temperature [e.g., Kato et al., 2004]. We use modeled temperatures and hydrostatic fluid pressures to determine fluid density and viscosity throughout the system. Over the range of estimated temperatures and pressures encountered, fluid density and viscosity are fairly insensitive to pressure [Parry et al., 2000]. Therefore the use of hydrostatic fluid pressures, rather than the spatially variable fluid overpressures likely present in subduction zones, has little effect on the determination of fluid properties. The fluid properties and permeability are used to determine the magnitude of trench-parallel differences in hydraulic conductivity for a range of differences in temperature. To quantify the differences in hydraulic conductivity, we use the ratio of hydraulic conductivity on the hot side of the system relative to that on the cold side.
We assign porosity within the wedge assuming an exponential decrease with depth [Athy, 1930; Bray and Karig, 1985]. Porosity of the décollement and underthrust sediment is set to be higher than in the wedge, to simulate underconsolidation within the underthrust section [e.g., Saffer and Bekins, 1998]. To define sediment permeability (k) in our models, we consider two end-member cases. In the first case, permeability varies only as a function of porosity (n):
where the coefficients in equation (3) are consistent with values compiled for argillaceous sediments [e.g., Neuzil, 1994]. In the second case, we specify that permeability decreases with temperature by decreasing k in equation (3) by one order of magnitude for every 220°C increase in temperature [Kato et al., 2004]. The permeability of the décollement zone is assigned a higher value, 10−14 m2 [Bekins et al., 1995]. For simulations with temperature-dependent permeability, décollement permeability is also reduced one order of magnitude for every 220°C increase in temperature.
4.4. Fluid Source Terms
As sediments are subjected to increasing stresses and temperatures due to burial and accretion, porosity loss and mineral dehydration release fluids. This dewatering causes both elevated pore pressures and resulting fluid flow. Porosity loss in the subducting sediment section provides the largest fluid source [e.g., Bekins et al., 1995; Saffer and Bekins, 1998; Screaton et al., 1990]. The magnitude of this fluid source tends to decrease with distance into a subduction zone, due to strain hardening of sediments with increasing effective stress and decreasing porosity, as documented by the typically observed exponential porosity decrease with depth [Athy, 1930]. The distribution of underthrust fluid sources with distance from the trench is controlled by burial rate (modulated by convergence rate and wedge taper angle) and initial porosity (controlled by the sediment type and the décollement depth at the deformation front). Rapid burial of the underthrust and subduction of shallow (high porosity) material result in large fluid sources concentrated near the trench and a dramatic decrease in fluid source magnitude with distance into the subduction zone. In contrast, slow burial and subduction of low porosity material result in fluid source magnitudes that decrease gradually.
In our models, we define compaction-driven fluid sources from estimates of porosity distribution in underthrust sediments for the Barbados, Nankai, and Costa Rica subduction zones [Bekins et al., 1995; Saffer and Bekins, 1998; Spinelli et al., 2006]. The primary trends in fluid source magnitude (Γ) with distance into the subduction zone (x) result from sediment compaction rates and are well approximated as exponential functions:
(Figure 8). In the Costa Rica underthrust section, the rapid burial (∼25 m/kyr) of shallow incoming sediment leads to large fluid source magnitude near the trench that decays dramatically with distance into the subduction zone (Figure 8). Lower fluid source magnitudes near the trench and more gradual decreases in source magnitude at Barbados and Nankai result from slower burial rates (∼1.5–2.9 m/kyr) of more deeply buried material on the incoming plate (i.e., a deeper décollement) and lower permeabilities that limit fluid expulsion. Small variations from the exponential trends are due to fluids from diagenetic reactions, rather than the main compaction driven source. The distribution of these diagenetic fluid sources depends on the thermal history of the sediment. Therefore differences in subduction zone temperature across a thermal transition will affect the distributions of these sources [e.g., Spinelli et al., 2006]. However, such variations in these source distributions across a thermal transition have only a small influence on the fluid pressure distribution within the system.
The distribution of fluid sources affects fluid pressure gradients; therefore fluid source distribution may also influence the relative importance of trench-perpendicular versus trench-parallel fluid flow. To explore this effect, we vary the distribution of fluid sources in the underthrust sediment in our simulations, but maintain a constant integrated fluid source over the system. The constant integrated source magnitude is averaged from the Barbados, Nankai, and Costa Rica trends from 0 to 80 km into a subduction zone. Steeper decreases in source magnitude with distance (more negative b in equation (4)) are compensated by larger sources in the shallow subduction zone (larger A); the total volume released from the underthrust to the system is the same for all simulations.
We use the décollement depth at the trench and underthrust sediment burial rate for the Chile, Mexico, and Cascadia sites to approximate the underthrust fluid source distribution relative to the trends for Costa Rica, Barbados, and Nankai (Table 1). Due to deep burial prior to subduction, the underthrust sediment on the Cascadia margin likely has low porosity at the trench, with little potential for additional porosity loss. On the basis of this and the modest burial rates, the decrease in fluid source magnitude with distance in the Cascadia subduction zone is likely no greater than on the Nankai margin. The depth to the décollement at the Chile and Mexico trenches is similar to that at the Nankai margin, but the underthrust sediment burial rates are higher. We use the decrease in underthrust source magnitude with distance for Nankai as a minimum estimate for Chile and Mexico. Underthrust sediment on the Chile and Mexico margins is buried at about one-half to one-third the rate of Costa Rica; we use one-half the decrease in underthrust source magnitude for Costa Rica as an upper limit for the Chile and Mexico margin estimates. More precise estimates of the underthrust fluid source distribution for these margins require estimates of rates of porosity loss and diagenetic reaction progress. Here, we are primarily interested in examining how the fluid source distribution and hydraulic conductivity contrast (both within ranges observed or estimated for subduction zones) affect the propensity for trench-parallel fluid flow.
Table 1. Controls on the Distribution of Fluid Sources in Underthrust
Wedge Taper at Trench
Convergence Rate, cm/yr
Underthrust Sediment Burial Rate, m/kyr
Décollement Depth at Trench, m
Γ (Vfluid Vsediment−1 s−1)
5.4 × 10−13e−0.073x
7.9 × 10−15e−0.019x
2.2 × 10−14e−0.0042x
Cascadia (Vancouver Is.)
Cascadia (Blanco FZ)
5.1. Subduction Zone Temperature and Hydraulic Conductivity
Décollement and sediment temperatures increase with distance into a subduction zone. At low temperatures and pressures, fluid viscosity is very sensitive to changes in temperature. At high temperature and pressure, fluid viscosity is much less sensitive to increasing temperature, so small decreases in density can counteract decreases in viscosity. Therefore, with constant permeability, the largest trench-parallel differences in hydraulic conductivity are close to the trench (i.e., in the coolest areas; Figure 9a). For the same permeability, the hydraulic conductivity in a warm subduction zone is higher than in a cool subduction zone, due to the dependence of hydraulic conductivity on fluid properties.
Including the effect of temperature-dependent permeability results in more complex differences in subduction zone hydraulic conductivity across a thermal transition. Elevated hydraulic conductivity on the hot side of the system is favored by low overall temperatures, because the difference in fluid viscosity is maximized and the effects of temperature on k are small (Figure 9b). For example, at the Costa Rica margin, décollement temperatures in the shallow subduction zone are low because the plate boundary is shallow (almost all the sediment at the trench is subducted). Thus, even a modest trench-parallel difference in temperature (20°C) results in a large difference in décollement hydraulic conductivity.
Because fluid viscosity is most sensitive to temperature changes at low temperatures (less than 100°C), with constant permeability the same trench-parallel difference in temperature between hot and cold portions of a subduction zone results in a larger difference in hydraulic conductivity in a system that is colder overall (Figure 10a). For example, at 50°C hydraulic conductivity is 1.58 times higher than at 25°C, but at 150°C hydraulic conductivity is only 1.18 times higher than at 125°C (Figure 10a). For the case with temperature-dependent permeability, at temperatures greater than ∼75°C, the hydraulic conductivity on the cold side of the system is higher than on the hot side; reduced permeability overcomes the accompanying decrease in fluid viscosity. For example, we estimate that the hydraulic conductivity in the shallow Cascadia (Blanco Fracture Zone and Vancouver Island) and Mexico subduction zones is ∼1.2–1.4 times higher on the cold side of the systems than on the hot side (Figure 10b). Near the trench in the Chile subduction zone, hydraulic conductivity is ∼10% higher on the cold side of the system than on the hot side. The difference in hydraulic conductivity in the décollement at Nankai Trough is small (1.03 times higher on the hot side than the cold side) due to the moderate temperatures and small temperature difference (Table 2).
Table 2. Trench-Parallel Differences in Subduction Zone Temperature and Hydraulic Conductivity, K
Cause of Trench-Parallel Temperature Difference
Décollement Temperature 5 km Landward of Trench on Cold Side, °C
Trench-Parallel Difference in Temperature, °C
Approximate Width of Thermal Transition at Trench, km
Chile (Guafo FZ)
Mexico (Rivera FZ)
Cascadia (Vancouver Is.)
Cascadia (Blanco FZ)
In addition to differences in the thermal state of incoming crust, the trench-parallel difference in subduction zone temperature is also affected by differences in convergence rate or subduction zone geometry along the trench. For example, in the Nankai subduction zone, the section off the axis of the fossil spreading center has relatively cool subducting crust and therefore lower shallow subduction zone temperatures than the section overlying the fossil spreading center. However, the section away from the fossil spreading center has a steeper wedge taper and its décollement temperature increases more rapidly with distance into the subduction zone. Thus this section is the “cold” section near the trench, but deep in the system its décollement is hotter than that of the section overlying the fossil spreading center. The side of the system defined as “hot” at the trench is colder beyond ∼60 km into the subduction zone. As a result the trend in the hydraulic conductivity ratio with distance for the Nankai example is small and reversed (Figure 9b).
5.2. Pore Pressures and Fluid Flow
Trench-parallel differences in subduction zone hydraulic conductivity cause trench-parallel fluid flow. Fluid driven out of dewatering sediment and exiting the system tends to flow from the low hydraulic conductivity side of the system to the high hydraulic conductivity side. We simulate fluid flow for subduction zones with a range of trends in hydraulic conductivity ratio (Khot/Kcold). The amount of trench-parallel fluid flow increases with increasing contrast in hydraulic conductivity across the thermal transition. As a measure of the amount of trench-parallel flow within the décollement, we show the flow line that separates fluid contained on one side of the system from fluid that crosses the center of the thermal transition (Figure 11).
Within the range of temperature driven hydraulic conductivity differences we examined, the trench-parallel component of fluid flow can dominate the flow regime (Figure 11b). However, the amount of trench-parallel fluid flow due to temperature differences is significantly modified by the fluid source distribution. Rapid decreases in fluid source magnitude with distance into the subduction zone result in a large trench-parallel component of fluid flow, whereas gradual decreases in fluid source magnitude with distance minimize the trench-parallel flow. A gradual decrease in source magnitude with distance results in large fluid sources that persist far into the subduction zone. In order to drive these fluids the long distance out of the system, large excess fluid pressures must develop. Therefore systems with gradual decreases in fluid source magnitude have larger fluid pressure gradients driving fluid perpendicular to the trench (i.e., out the décollement to the trench or up through the wedge to the seafloor) than systems with rapid decreases in fluid source magnitude near the trench. For the same trench-parallel difference in hydraulic conductivity, small trench-parallel differences in fluid pressure may be an important control on fluid flow patterns where fluid sources decrease dramatically with distance, yet they are mostly overwhelmed by large trench-perpendicular pressure gradients where fluid sources are more uniformly distributed. The change in fluid source magnitude with distance controls the pattern of fluid pressures and fluid flow; varying the total integrated fluid source in the system changes the magnitude of fluid pressures throughout the system, but has little effect on the fluid pressure pattern.
6. Discussion and Implications
6.1. Application to Active Margins
The thermal conditions and fluid source distributions at Costa Rica and the Cascadia margins provide essentially opposite end-members regarding trench-parallel fluid flow. The shallow décollement on the Costa Rica margin (nearly all incoming sediment is subducted) is very cold (∼2–15°C), allowing the development of large trench-parallel differences in fluid viscosity due to modest temperature differences. Additionally, the rapid burial of shallow, high porosity sediment generates large fluid sources near the trench that diminish in magnitude rapidly with distance into the subduction zone. As a result, the Costa Rica margin supports a large component of fluid flow from the cold side of the system to the hot side (Figure 12a). These results suggest that fluid flowing in the décollement near the thermal transition at the trench (e.g., near the Ocean Drilling Program Leg 170 sites) may have traveled nearly as far parallel to the trench as perpendicular to it.
In contrast, the thick sediment section above the décollement at the Cascadia trench results in a hot décollement (120–230°C). Trench-parallel differences in hydraulic conductivity are dominated by differences in permeability rather than fluid viscosity. Fluid flow is from the hot side of the margin to the cold side. In addition, subduction of highly compacted, low porosity sediment likely results in a very gradual decrease in fluid source magnitude with distance. Therefore, despite the larger overall trench-parallel difference in temperatures at Cascadia than Costa Rica, the trench-parallel component of fluid flow is more modest due to the gradual decrease in fluid source magnitude (Figure 12c).
Unlike either the Cascadia or Costa Rica margins, the Nankai margin has a low potential for trench-parallel fluid flow. The small difference in hydraulic conductivity and very gradual decrease in fluid source magnitude with distance result in almost no trench-parallel component of fluid flow (Figure 12c).
The large contrasts in hydraulic conductivity and moderate decreases in estimated underthrust fluid source magnitude on the Chile and Mexico margins make them more amenable to supporting thermally driven trench-parallel fluid flow than the Cascadia or Nankai margins (Figure 12b). However, the wide thermal transitions at Chile and Mexico tend to reduce trench-parallel pressure gradients. Therefore the flow line to the center of the thermal transition at these margins should be shifted closer to the thermal transition than illustrated in Figure 12 for a subduction zone with a 5 km wide thermal transition. For the Chile margin, the flow line to the center of a 5 km wide thermal transition extends between 23 and 58 km onto the hot side of the system at 80 km into the subduction zone (Figure 12); for the same conditions, the flow line to the center of a 100 km wide thermal transition extends between 18 and 54 km onto the hot side at 80 km into the subduction zone.
6.2. Fluid Pressure and Mechanical Implications
In all of our simulations, modeled fluid pressure in the décollement is higher on the low hydraulic conductivity side of the thermal transition than on the high conductivity side (e.g., Figure 13). This difference in fluid pressure drives the fluid to the high conductivity side of the system. For cold systems, fluid pressure is higher on the cold side of the system because of the higher fluid viscosity there; for hot systems, fluid pressure is higher on the hot side of the system due to reduced permeability. Trench-parallel changes in fault strength may be related to differences in fluid pressure, because fluid pressure controls effective stress. Effective stress on the low conductivity side of the system is reduced commensurate with the higher fluid pressures. Along the plate boundary fault, reduced effective stress on the low conductivity side of the system relative to the high conductivity side will result in a reduced effective friction coefficient. Since increasing effective stress is linked to the onset of frictional instability, trench-parallel variations in fluid pressure may also affect the position of the updip limit of seismicity [Moore and Saffer, 2001; Scholz, 1998].
For a cold system (e.g., Costa Rica), a series of friction-related positive feedbacks may enhance and maintain trench-parallel temperature and pressure differences. Low fluid pressure and high effective stress on the hot side of the system (Figure 13) should result in enhanced frictional heating on the hot side relative to the cold side. Additionally, the cementation and lithification processes that reduce permeability with increasing temperature probably also increase the friction coefficient at a given depth on the hot side relative to cold side. These feedbacks may amplify the pore pressure difference in cold systems. In contrast, hot systems with reduced fluid pressure and elevated effective stress on the cold side of the system will have a negative feedback due to increased frictional heating on the cold side.
7. Summary and Conclusions
Trench-parallel differences in subduction zone temperature can be caused by hydrothermal circulation, local thermal rejuvenation of the incoming plate by volcanism, or differences in plate age. These differences in subduction zone temperature can result in trench-parallel differences in fluid pressure. Such differences in fluid pressure control effective stress within the plate boundary fault system; in turn, they should affect fault strength and may relate to the updip limit of seismicity. Additionally, trench-parallel differences in fluid pressure within the plate boundary fault zone drive trench-parallel fluid flow and thus hold important implications for the origin and flow path of pore fluids sampled near the trench or on the continental slope. Because temperatures along subduction zone megathrusts generally span a range of temperatures over which fluid viscosity and sediment permeability change substantially, such trench-parallel variability may be important in many subduction zones characterized by trench-parallel differences in the thermal state of the incoming plate. Substantial trench-parallel flow is favored by large temperature contrasts and rapid decreases in underthrust fluid source magnitude. Where overall subduction zone temperatures are low, flow from the cold side of a system to the hot side is favored. Flow from the hot side to the cold side is favored in systems with higher temperatures.
Settings well suited for large trench-parallel differences in the thermal state of subducting crust include: young (i.e., rapidly cooling) plates with trench-parallel differences in age, and young (i.e., warm) plates with vigorous and patchy hydrothermal circulation. Hydrothermal circulation systems are likely present within most ocean crust; this is likely the most common cause of local variations in the thermal state of subducting crust. On the basis of global heat flow anomalies, circulation systems that extract a measurable quantity of lithospheric heat from the crust (i.e., open circulation systems [Jacobson, 1992]) are present in crust up to ∼65 Ma in age [Anderson et al., 1977; Fisher, 2005; Stein et al., 1995]. Whether hydrothermal circulation within crust approaching a subduction zone is open or closed (i.e., redistributing heat locally, but not extracting heat from the crust) depends largely on the extent and distribution of low permeability sediment capping the fractured basaltic basement rock [Spinelli et al., 2004]. Detailed studies of the thermal state of crust approaching subduction zones may reveal many locations where trench-parallel changes in temperature influence décollement fluid pressures, fluid flow patterns, and the strength of plate boundary faults.
This research was supported by a JOI/USSAC grant to Saffer and NSF grant OCE-0304946. We thank two anonymous reviewers and associate editor Eli Silver for helpful comments on this manuscript.