Volume 124, Issue 2 p. 1205-1221
Research Article
Free Access

The Thermal Effects of Plate-Bending-Related Thickening of the Oceanic Crustal Aquifer in the Nankai Trough and Japan Trench Subduction Zones

A. C. Lucero

Corresponding Author

A. C. Lucero

Department of Earth and Environmental Science, New Mexico Institute of Mining and Technology, Socorro, NM, USA

Correspondence to: A. C. Lucero,

[email protected]

Search for more papers by this author
G. A. Spinelli

G. A. Spinelli

Department of Earth and Environmental Science, New Mexico Institute of Mining and Technology, Socorro, NM, USA

Search for more papers by this author
J. He

J. He

Pacific Geoscience Centre, Geological Survey of Canada, Sidney, British Columbia, Canada

Search for more papers by this author
First published: 08 February 2019
Citations: 1

Abstract

For the Nankai Trough and Japan Trench subduction zones, we examine the combined thermal effects of lateral heat exchange by fluid circulation in an oceanic crustal aquifer and the thickening of that aquifer due to plate-bending-related faulting. Faults induced by the bending of a plate entering a subduction zone are hypothesized to increase the depth over which vigorous hydrothermal circulation can redistribute heat in the oceanic crust. Previous 1-D (vertical) thermal models have examined how aquifer thickening can mine heat from deep in the crust seaward of the trench. Here we construct 2-D thermal models that include aquifer thickening and lateral heat exchange in the aquifer. We vary the maximum aquifer thickness and the landward extent to which hydrothermal circulation persists within the subducted crust. For the Nankai margin, models most consistent with heat flux data require vigorous fluid circulation extending up to 150 km landward of the trench; aquifer thickening is permitted but not required for models to be consistent with the heat flux observations. Conversely, for the Japan Trench, preferred models include substantial aquifer thickening (maximum aquifer thickness of 1.8–5 km); vigorous circulation extending landward of the trench is permitted but is not required. For hot subduction zones, the thermal effects of aquifer thickening are modest relative to the large lateral advective heat redistribution. For cold subduction zones, where small lateral temperature gradients limit the amount of lateral heat redistribution, aquifer thickening can be the dominant process generating thermal anomalies.

Key Points

  • Lateral heat exchange by fluid circulation in a crustal aquifer and faulting-induced aquifer thickening affect subduction zone temperatures
  • For hot subduction zones, the thermal effects of aquifer thickening are modest relative to the large lateral advective heat redistribution
  • For cold subduction zones, aquifer thickening can be the dominant process generating thermal anomalies

1 Introduction

Temperatures in subduction zones are important controls on the alteration and dehydration of subducting material, frictional behavior, mantle wedge hydration, and melt generation (e.g., Hacker et al., 2003; Peacock, 2009; van Keken & King, 2005; Wada et al., 2012). Plate convergence rate, subducting plate age, and slab dip are basic factors affecting subduction zone temperatures (e.g., Dumitru, 1991). Anomalously high surface heat flux along the Nankai Trough and Japan Trench (Yamano et al., 2003, 2014), relative to expectations for conductively cooling oceanic lithosphere (e.g., Stein & Stein, 1992), suggests the influence of a regional heat transport mechanism in addition to heat conduction (e.g., Harris et al., 2017). Several previous thermal models of these regions produce higher heat flux near the trench by including the effects of hydrothermal circulation in an aquifer in the upper basaltic basement of the oceanic crust where permeability is high (e.g., Kawada et al., 2014; Spinelli, 2014; Spinelli & Wang, 2008). In one subset of these thermal models, the key function of hydrothermal circulation is to redistribute heat laterally between hot subducted crust and cooler crust seaward of the subduction zone (e.g., Harris et al., 2013; Spinelli & Wang, 2008). Another set of studies examines the thermal effects of fluid circulation as the aquifer thickens due to plate bending near the outer rise; this mines heat from deeper in the crust (e.g. Kawada et al., 2014; Yamano et al., 2014). However, neither of these subsets of models include the effects of both aquifer thickening and lateral heat exchange in the subducting aquifer. The purpose of this study is to compare a suite of thermal models with different distributions of vigorous hydrothermal circulation in the subducting plate. We do this by considering three contrasting scenarios: (1) lateral heat exchange in a constant thickness aquifer that crosses the trench from the subducted crust to the crust seaward of the trench, (2) lateral heat exchange in an aquifer that thickens prior to subduction that is limited to the crust seaward of the trench, and (3) both continued lateral heat exchange landward of the trench and aquifer thickening prior to subduction (Figure 1).

Details are in the caption following the image
Conceptual models for three hypothetical distributions of vigorous hydrothermal circulation in subducting crust. Crosshatched areas indicate portions of the system with vigorous fluid circulation redistributing heat. (a) Lateral redistribution of heat in aquifer of constant thickness crossing the trench. (b) Aquifer thickening prior to subduction due to plate-bending mining heat from lower basement. (c) Both aquifer thickening and continued circulation in the subducting plate.

Fluid circulation in the oceanic crust redistributes heat, affecting patterns of surface heat flux (e.g., Davis & Lister, 1977; Davis et al., 1999; Fisher et al., 2003; Lister, 1972). In the oceanic crust, vigorous fluid circulation is limited to a high-permeability aquifer commonly described as comprising the upper ∼600 m of basaltic basement rock (Fisher, 1998). Fractures, pillow basalt breccias, and the contacts between individual basalt flows in these rocks result in the aquifer having very high permeability (i.e., up to  10–9 m2; e.g., Fisher et al., 2014). The oceanic crustal aquifer is confined above and below by low-permeability sediment and intrusive rocks, respectively (e.g., Fisher, 1998). Fluid circulation in the highly permeable aquifer redistributes heat and homogenizes temperatures in the upper oceanic basement rocks (e.g., Davis et al., 1997). In subduction zones, the high permeability of the aquifer can be maintained after subduction, thus facilitating fluid circulation that mines heat from the subducted crust and transports it updip to the crust seaward of the trench (e.g., Harris et al., 2017; Kummer & Spinelli, 2008; Spinelli & Wang, 2008). Lateral heat exchange in the subducting crust is driven by the temperature gradient between the subducted crust and the incoming crust (Rotman & Spinelli, 2013). Thus, the thermal effects of hydrothermal circulation in subducting crust are likely more pronounced for hotter subduction zones (Rotman & Spinelli, 2013).

As a subducting plate traverses the outer rise, tension in the upper part of the plate causes normal faulting (Ludwig et al., 1966), which could extend deep into the crust (e.g., Ranero et al., 2003), enhance permeability, and effectively increase the thickness of the oceanic crustal aquifer (Kawada et al., 2014). Increasing aquifer thickness could affect the depth over which fluid circulation can redistribute heat (Kawada et al., 2014; Yamano et al., 2014). Plate-bending normal faults can extend the depth to which thermally significant fluid circulation occurs in the crust (e.g., Ranero et al., 2003), thereby incorporating heat from deeper in the basement into the aquifer. The increased temperature in the aquifer increases the conductive heat flux from the top of the aquifer, through the overriding sediment, to the seafloor. The thermal effects of plate-bending-related aquifer thickening are expected to be greater for slabs with a larger degree of curvature. A greater degree of curvature is associated with more tension at the top of the slab, thus resulting in more extensive bend-related normal faulting than in areas with a small degree of slab curvature (Faccenda, 2014). For example, anomalously low P wave velocities in the crust approaching the Japan Trench, the Middle America trench, the Chile trench, and the Tonga trench are interpreted as an indication of pervasive bending-related faulting and fracturing (Contreras-Reyes et al., 2008, 2011; Ivandic et al., 2008; Kawada et al., 2014; Ranero & Sallarès, 2004); the degree of plate hydration appears to correlate with the amount of slab bending (Faccenda, 2014). Previous thermal models considering fluid circulation in subducting crust used a fixed-thickness aquifer (e.g., Rotman & Spinelli, 2014; Spinelli & Wang, 2008). This study explores the effects of both aquifer thickening prior to subduction and continued hydrothermal circulation landward of the trench, simultaneously.

In this study, we examine heat transport in the Nankai Trough and the Japan Trench subduction zones. We examine the Nankai Trough along two transects; one offshore Cape Muroto and one through the Kumano Basin (Figure 2). The Nankai Trough is located off the coast of southern Japan, where the Philippine Sea Plate subducts beneath the Amurian Plate. For this relatively young lithosphere (15–20 Ma), surface heat flux for conductively cooled lithosphere is expected to be ∼112–130 mW/m2 (Stein & Stein, 1992). The Nankai Trough is one of the most studied subduction zones, including extensive seafloor drilling activities. Heat flux, thermal conductivity, and radiogenic heat production data were gathered from marine boreholes through the Deep Sea Drilling Project, the Ocean Drilling Program, and the Integrated Ocean Drilling Program (Fulton et al., 2013; Harris et al., 2013; Moore et al., 2001; Shipboard Scientific Party, 2000). Additional heat flux measurements were gathered using gravity-driven probes (Kinoshita & Yamano, 1986; Watanabe et al., 1970; Yamano et al., 1984, 1992, 2003).

Details are in the caption following the image
Location maps showing heat flux measurements along (a) the Nankai Trough and (b) the Japan Trench. Inset map of Japan shows the location of the study areas. Circles indicate marine probe data (Hamamoto et al., 2011; Kinoshita & Yamano, 1986; Mikada et al., 2003; Yamano et al., 2003, 2014), triangles indicate borehole observations (Fulton et al., 2013; Moore et al., 2001; Shipboard Scientific Party, 2000), and stars indicate long-term temperature monitoring sites on the continental shelf (Hamamoto et al., 2005, 2011).

A first-order feature in the suite of heat flux observations for the Nankai margin is that the area near the trench is anomalously hot (Watanabe et al., 1970; Yamano et al., 1984, 2003). Heat flux observations more than 30 km seaward of the trench average ∼20% below the expectation for conductively cooled oceanic lithosphere and are consistent with the global average for ∼15- to 20-million-year-old lithosphere (likely indicative of ventilated hydrothermal circulation extracting heat from the lithosphere; Harris et al., 2013). Within 30 km of the trench axis, surface heat flux measurements on the Philippine Sea plate average 20% higher than expected for conductively cooled lithosphere (Harris et al., 2013). Landward of the trench, surface heat flux decreases more rapidly than expected based on the cooling effects of the subducting plate and the landward thickening of the margin wedge (Wang et al., 1995). These heat flux anomalies extend for at least 300 km along the strike of the margin, suggesting a regional mechanism as their cause. The juxtaposition of anomalously low heat flux >30 km from the trench with anomalously high heat flux within 30 km of the trench requires an input of heat near the subduction zone (Harris et al., 2013). Spinelli and Wang (2008) attribute these thermal anomalies to hydrothermal circulation in the subducting aquifer transporting heat from the subducted crust seaward to the incoming plate.

The Japan Trench is located off the northeast coast of Japan where the Pacific Plate subducts beneath the Okhotsk Plate. The age of the Pacific Plate in this area is ∼135 Ma (Müller et al., 2008); the expected surface heat flux for a conductively cooling plate of this age is ∼50 mW/m2 (Stein & Stein, 1992). Observations along three transects show surface heat flux consistent with expectations for conductively cooled lithosphere on seafloor >150 km seaward from the trench (Figure 2; Yamano et al., 2014). On crust <150 km from the trench, the average observed heat flux is >2.5 times the expectation for conductively cooled oceanic lithosphere (Yamano et al., 2014). These thermal anomalies are broad (∼100 km wide) and span the outer rise of the subduction zone (Yamano et al., 2014).

Kawada et al. (2014) hypothesize that the thermal anomalies offshore from the Japan Trench result from the oceanic crustal aquifer thickening as the plate traverses the outer rise. In this scenario, plate-bending normal faults increase the thickness of the zone hosting vigorous fluid circulation and thus influence heat redistribution in the system. One indication that bend-related normal faulting may affect the depth extent of fluid circulation in this area comes from anomalies in Vp/Vs ratios. In a transect perpendicular to the Japan Trench, Vp/Vs in the upper 2 km of oceanic crust is nearly constant (∼1.85) seaward of the outer rise (Fujie et al., 2013). From the outer rise to the trench, Vp/Vs in the upper 2 km of the crust is higher (∼1.90; Fujie et al., 2013). Vp/Vs increases with increasing water content if the crack aspect ratio is <0.03 (i.e., for thin cracks) (Takei, 2002). Fujie et al. (2013) hypothesize that bending-related stress opens fractures in the upper crust, thus increasing porosity and increasing Vp/Vs. In addition, multichannel seismic reflection profiles along transects perpendicular to the Japan Trench show surficial evidence of fracturing in the form of horst and graben structures and normal faults within 80 km of the trench (Ito et al., 2004; Kobayashi et al., 1998; Miura et al., 2005; Tsuru et al., 2000). Kawada et al. (2014) suggest that the opening of fractures in the oceanic crust increases the thickness of the aquifer that can host vigorous hydrothermal circulation. In considering the thermal effects of aquifer thickening, previous studies treat the aquifer seaward of the trench as completely isolated from the aquifer in the subducted plate and do not consider possible effects of aquifer thinning as the plate unbends (Kawada et al., 2014; Ray et al., 2015). Here we examine the combined thermal effects of aquifer thickening due to plate-bending normal faults and lateral heat exchange in a subducting aquifer.

In this study, we examine the combined thermal effects of aquifer thickening due to plate-bending normal faults and lateral heat exchange in a subducting aquifer for both the Nankai Trough and the Japan Trench subduction zones. The thermal effects of aquifer thickening may be greater in the Japan Trench relative to the Nankai Trough due to its greater degree of slab curvature, which increases the depth from which heat can be incorporated into the aquifer by vigorous fluid circulation. Relative to the Philippine Sea Plate along the Nankai Trough, the Pacific Plate at the Japan Trench is older (∼135 Ma vs. 15–25 Ma; Müller et al., 2008) and thicker (82 km vs. 55 km; Kawakatsu et al., 2009) and dips more steeply upon subduction (29–34° vs. 20–31°; Hayes et al., 2012). For the Pacific Plate approaching the Japan Trench, there is a prominent outer rise with its crest ∼150 km seaward of the trench. The crest of the outer rise in this area averages ∼170 m higher than the seafloor 100 km farther seaward (Ryan et al., 2009). In contrast, the younger and thinner Philippine Sea Plate approaching the Nankai Trough has no bathymetric expression of an outer rise. Rather, from 250–50 km seaward of the trench, the Philippine Sea plate dips gently (∼0.002°) toward the trench; at ∼50 km seaward of the trench, the dip of the seafloor toward the trench increases to ∼0.5° (Ryan et al., 2009). Because fluid circulation is more effective at laterally redistributing heat in hotter subduction zones (Rotman and Spinelli, 2013), the thermal effects of that lateral heat exchange may be greater in the Nankai Trough (relative to the Japan Trench).

2 Methods

We use the 2-D finite element model PGCtherm2D, written by Jiangheng He, to simulate subduction zone temperatures. The model has been benchmarked (van Keken et al., 2008) and has been applied extensively (e.g., Perry et al., 2016; Völker et al., 2011; Wada & Wang, 2009; Wada et al., 2012). The base model consists of a subducting plate with prescribed motion, fixed upper plate, and viscous mantle wedge. Heat is produced by fault friction and radioactive decay. Heat is transported by conduction and advection of the subducting slab and mantle wedge flow. We consider two transects perpendicular to the Nankai Trough and three transects perpendicular to the Japan Trench. The modeled geometry for each transect is constrained by gravity and seismic data (Baba et al., 2002; Hirose et al., 2008; Nakajima et al., 2001; Nakanishi et al., 2002; Park et al., 2002; Tobin et al., 2015). The land surface and seafloor are the top of the domain; the bottom of the model is 100 km below the top of the subducting slab. The oceanic lithosphere consists of sediment, a crustal aquifer, and lower basement rocks. The upper portion of the domain (i.e., the top of the slab, the mantle wedge, and the upper plate) extend 400 km landward of the trench. For the Nankai Trough, the top of the slab extends 100 km seaward of the trench; for the Japan Trench, the top of the slab extends 250 km seaward of the trench. Models for the Japan Trench extend farther seaward (relative to those for the Nankai margin) because available heat flux data extends farther seaward offshore from the Japan Trench, and to account for the outer rise being far from the trench for the relatively old Pacific Plate in this area.

We use a method developed by Spinelli and Wang (2008) to simulate vigorous hydrothermal circulation in the oceanic crustal aquifer. Since achieving numerical stability can be difficult when simulating coupled heat and fluid transport in an extremely permeable aquifer, high thermal conductivity in the basement aquifer is used as a proxy for vigorous fluid circulation (Davis et al., 1997). High thermal conductivity in the aquifer simulates the thermal effects of vigorous fluid flow without knowledge of the exact nature of flow patterns within the aquifer. The Rayleigh number for aquifer elements is calculated using the equation
urn:x-wiley:jgrb:media:jgrb53301:jgrb53301-math-0001(1)
where g is acceleration due to gravity, k is permeability, L is the aquifer thickness, K is the thermal conductivity, and κ is the thermal diffusivity (Spinelli & Wang, 2008). Fluid thermal expansivity (α), density (ρ), and viscosity (μ) are determined using the pressure and temperature of each aquifer element (Harvey et al., 2014). The conductive heat flux into the base of the aquifer is determined using the vertical temperature gradient at the base of each aquifer element. The aquifer permeability is set to 10−9 m2 prior to subduction and decreases logarithmically by 0.015 per kilometer of burial depth (e.g., Rotman & Spinelli, 2013). The Nusselt number is then calculated using the empirical equation from Spinelli and Wang (2008):
urn:x-wiley:jgrb:media:jgrb53301:jgrb53301-math-0002(2)

For each aquifer element, the intrinsic thermal conductivity of the aquifer is multiplied by the Nusselt number; then, the conductive model is rerun. The resulting fluid properties are used to recalculate the Rayleigh numbers and then update the Nusselt numbers for each aquifer element. This process is repeated until the temperatures stabilize.

We examine a range of distributions of hydrothermal circulation in the subducting plate. We use the method described above to define the thermal conductivity for aquifer elements in the region where vigorous fluid circulation occurs; for aquifer elements where vigorous fluid circulation does not occur (i.e., those below or landward of the defined region with vigorous fluid circulation) in a particular simulation, we use the intrinsic thermal conductivity for the basaltic basement rocks (Table 1). For most transects, we examine scenarios with the effects of vigorous fluid circulation in the aquifer terminating at the trench, and at 5, 15, 25, 35, 45, 75, and 100 km landward of the trench. This range of values for the landward extent of vigorous circulation captures simulations that minimize the misfit between modeled and observed surface heat flux for all the transects except Muroto. For the Muroto transect, we examine scenarios with vigorous circulation terminating at the trench, and at 50, 100, 150, 175, and 200 km landward of the trench. We do not allow vigorous fluid circulation (i.e., high thermal conductivity) where aquifer temperatures are greater than 800 ° C, the brittle-ductile transition for mid-ocean ridge basalt (Violay et al., 2015). At this temperature, high permeability associated with open fractures becomes unlikely.

Table 1. Thermal Properties of Stratigraphic Layers
Thermal Radiogenic heat
conductivity, production,
K (Wm-1 K-1) H (μWm-3) References
Sediment 1.10a, 1.0b 1.5 Marcaillou et al. (2012); Fulton et al. (2013);
Yamano et al. (2014); Sugihara et al. (2014)
Basaltic aquifer 3.1 Nuc 0.02 Wada and Wang (2009);
Harris et al. (2011); Spinelli and Harris (2011)
Lower oceanic lithosphere 3.1 0.02 Wada and Wang (2009); Harris et al. (2011); Spinelli and Harris (2011)
Upper continental crust 2.9a, 2.5b 1.8a, 1.3b Hyndman et al. (1995); Wada and Wang (2009); Harris et al. (2011);
Spinelli and Harris (2011); Marcaillou et al. (2012)
Lower continental crust 2.9a, 2.5b 0.4 Hyndman et al. (1995); Wang et al. (1995) Wada and Wang (2009); Harris et al. (2011);
Spinelli and Harris (2011); Marcaillou et al. (2012)
Continental mantle 3.1 0.02 Wada and Wang (2009);
Harris et al. (2011); Spinelli and Harris (2011)
  • a Nankai Trough.
  • b Japan Trench.
  • c Nu is the Nusselt number in the aquifer.

We simulate varying the thicknesses of the subducting aquifer by changing the portion of the model with high thermal conductivity. We run one set of simulations with an aquifer with constant thickness (0.6 km). In addition, we test a suite of models in which the maximum thickness of the subducting aquifer is increased to 1.2, 1.8, 3, and 5 km. The largest of the values considered (5 km) is near the full crustal thickness for crust generated at intermediate spreading rates (Bown & White, 1994). For the Japan Trench transects, the onset of aquifer thickening in our models occurs at 150 km seaward of the trench, coinciding with the crest of the outer rise, the seaward end of anomalous Vp/Vs in the oceanic crust (Fujie et al., 2013), and the seaward end of anomalously high surface heat flux (Kawada et al., 2014; Yamano et al., 2014). For these transects, we place the landward end of the thickened aquifer (i.e., where the aquifer returns to 0.6 km thick) at 255 km landward of the trench, the location at which the subducting slab has unbent (i.e., landward of here the slab top is straight; Nakajima et al., 2001). We set the maximum aquifer thickness at 45 km landward of the trench (i.e., the midpoint between 150 km seaward of the trench and 255 km landward of the trench). Aquifer thickening and thinning are linear and symmetrical about the midpoint. For the transects through the Nankai Trough, we place the onset of aquifer thickening in our models at 50 km seaward of the trench, coinciding with an increase in the seafloor dip toward the trench (there is no bathymetric indication of an outer rise seaward of the Nankai Trough). We set the maximum aquifer thickness at 45 km landward of the trench, consistent with the models for the Japan Trench. As with the Japan Trench models, we impose symmetrical aquifer thickening and thinning; the landward end of the thickened aquifer Nankai Trough transects is at 140 km landward of the trench.

The final aspect of heat redistribution by fluid circulation that we examine is the possible effect of anisotropic heat transport within the oceanic crustal aquifer. We consider scenarios in which vertical heat transport in the aquifer is enhanced but lateral heat transport in the aquifer is not. This could result from bend-related normal faults simultaneously increasing the vertical heat exchange between the lower and upper oceanic crust and disrupting lateral fluid flow in the upper oceanic crustal aquifer. Seismic reflection images along the transects through the Japan Trench reveal normal faults with vertical offsets >600 m generating horst and graben structures (Kimura et al., 2012), which could laterally compartmentalize the upper oceanic crustal aquifer. For these scenarios, we calculate Nu for aquifer elements as described above; the vertical thermal conductivity for the aquifer elements is multiplied by Nu, but the horizontal thermal conductivity for the aquifer elements remains at the intrinsic conductivity for the oceanic crust.

The boundary conditions for the thermal model include oceanic and continental geotherms, and constant temperature at the seafloor (0 °C) and at the base of the subducting slab (1,470 °C). For a reference simulation in which there is no hydrothermal circulation, the seaward geotherm is calculated based on conductive cooling of the lithosphere and sediment accumulation (Wang & Davis, 1992). Simulations including hydrothermal circulation utilize an isothermal oceanic aquifer at the seaward boundary; the temperature for the full thickness of the aquifer at the seaward boundary is equal to the temperature at the base of the sediment layer. Below the isothermal aquifer, the geotherm is calculated based on the temperature at the base of the aquifer and conductively cooled lithosphere. The landward geotherm is calculated from a typical surface heat flux for backarcs, 80 mW/m2 (Currie & Hyndman, 2006), and the material properties of the overriding plate (Table 1).

Simulations of the Japan Trench transects are run to steady state. However, simulations for the Nankai Trough are transient in order to capture the thermal effects of variations in convergence rate and the age of the subducting plate. We use the plate age and convergence rate for three time periods, beginning 10 million years ago (Mahony et al., 2011; Table 2). Initial temperature for the model at 10 million years before present is set from a model run to stead state with the plate ages and convergence rate for the time period from 13 to 10 Ma (Table 2). The modeled temperature distribution at the end of each time interval is used as the initial temperature distribution for the following time period. Over the last 10 Ma, the Philippine Sea Plate traveled >450 km landward (i.e., greater than the extent of the model landward of the trench).

Table 2. Plate Age and Convergence Rate for Nankai Trough Transects (After Mahony et al., 2011)
Time before present Age of subducting lithosphere (Ma) Convergence rate perpendicular to
(Ma) Muroto transect Kumano transect trench (cm/year)
13 → 10 10 20 5.14
10 → 5 16 11 4.30
5 → 2 16 13 4.60
2 → 0 15 20 4.91

3 Results

We consider a suite of models for each transect in which the landward extent of vigorous fluid circulation and the maximum aquifer thickness are varied. The results from these simulations are compared to local surface heat flux observations. For each transect, a conductive case in which there is no redistribution of heat by vigorous fluid circulation is used as a reference. Due to the scattered nature of the available surface heat flux data, we determine spatially binned (15 km) median heat flux values from the observations along each transect. The misfit between the modeled simulation and the heat flux data is determined by calculating the root mean square error (RMSE) between the modeled heat flux and the median observed heat flux in each bin.

Two transects in the Nankai margin are examined, one offshore Cape Muroto (through Ocean Drilling Program sites 808, 1173, and 1174) and one through the Kumano Basin (and through the Integrated Ocean Drilling Program Sites that are part of the Nankai Trough Seismogenic Zone Experiment; Figure 2). For the Muroto transect, the reference simulation does not have a high heat flux anomaly near the trench; modeled heat flux is up to 150 mW/m2 lower than the observations (Figure 3a; dashed line). The model that is most consistent with the observed heat flux values includes an aquifer that facilitates continued vigorous fluid circulation within the crust after subduction (Figure 3a; bold black line); the aquifer is 0.6 km thick (i.e., no aquifer thickening) and vigorous heat redistribution in the aquifer extends 150 km landward of the trench (Figure 3b). The lateral heat exchange in the aquifer increases heat flux (relative to the reference simulation) on the incoming plate and the seaward-most 40 km of the margin wedge and decreases heat flux >40 km landward of the trench (Figure 3a). In this simulation, the maximum amplitude of the thermal anomaly is 70 mW/m2 above the modeled heat flux in the reference simulation. Relative to the reference simulation, this preferred simulation reduces the RMSE between the modeled and observed heat flux by 21% (Figure 3b, star). If vigorous fluid circulation in a constant thickness (0.6 km thick) aquifer stops at the trench, the resulting thermal anomaly has a maximum amplitude of only 20 mW/m2 (Figure 3a, dark gray line); this is less consistent with the heat flux observations than the reference simulation, increasing the misfit between the modeled and observed heat flux by 4% (Figure 3b, dark square). Relative to the preferred simulation, increasing the maximum aquifer thickness to 5 km increases the maximum amplitude of the heat flux anomaly in the trench to 122 mW/m2 above the heat flux in the reference simulation (Figure 3a, light gray line). In this case with the maximum increase in aquifer thickness, the RMSE between the modeled and observed heat flux is increased by 13% from the reference simulation (Figure 3b, light square).

Details are in the caption following the image
(a) Measured and modeled surface heat flux along the Muroto transect of the Nankai margin; heat flux observations within 100 km of the transect are shown. The red squares are the median of heat flux observations in 15-km-wide bins. Gray circles, triangles, and stars are probe, bottom-simulating reflector (BSR), borehole, and long-term monitoring data, respectively. In the legend, HC indicates how far landward of the trench hydrothermal circulation in the aquifer persists; AQ indicates the maximum aquifer thickness. (b) Percent improvement in the misfit between modeled and observed heat flux from simulations with hydrothermal circulation relative to the reference simulation with no hydrothermal circulation, as functions of the maximum aquifer thickness and the landward extend of hydrothermal circulation. A positive value indicates the modeled heat flux for a given simulation fits the data better than the reference simulation. A negative value indicates that the modeled heat flux from the reference simulation fits the data better than the given simulation does. The star shows the preferred simulation (bold black line in panel a). The light gray square is a simulation with the same landward extent of circulation as the preferred simulation, but greater aquifer thickening (light gray line in panel a). The dark gray square is a simulation with the same aquifer thickness as the preferred simulation, but a smaller landward extent of circulation (dark gray line in panel a). Black circles indicate other simulations in this study.

For the Kumano transect through the Nankai Trough, the reference simulation with no hydrothermal circulation (Figure 4a, dashed line) has modeled heat flux that is lower than the average of the observations from 75 km seaward of the trench to 25 km landward of the trench. In this case, the modeled heat flux is higher than the average of the observations from 25–90 km landward of the trench (Figure 4a). However, the heat flux observations within 100 km of the transect (i.e., those shown in Figure 4a) are more scattered than those along the Muroto transect. Calculating the RMSE using 15-km binned data reduces some error associated with scatter, while capturing the general trend of the observational data. Relative to the reference simulation, a simulation with a 0.6-km-thick aquifer and vigorous lateral heat exchange in the aquifer extending 35 km landward of the trench reduces the misfit between the modeled and observed heat flux by 4% (Figure 4b, star). In this simulation, the maximum anomaly in modeled heat flux (58 mW/m2 higher than in the reference simulation) occurs at 43 km seaward of the trench (Figure 4a, bold black line). For all the results shown, the modeled heat flux landward of the trench is within 10/mW m2 of those from the reference simulation; larger differences in the modeled heat flux between simulations are restricted to 10–60 km seaward of the trench (Figure 4a). Having vigorous fluid circulation in a constant thickness (0.6 km thick) aquifer that continue for 100 km landward of the trench results in modeled heat flux deviating from the preferred simulation by <25 mW/m2 (Figure 4a, dark gray line). This is slightly less consistent with the heat flux observation than the preferred simulation; the misfit between the modeled and observed heat flux is 1% larger than for the reference simulation (Figure 4b, dark square). Relative to the preferred simulation, increasing the maximum aquifer thickness to 5 km increases the amplitude of the heat flux anomaly 43 km seaward of the trench to 143 mW/m2 (Figure 4a, light gray line). In this case with the maximum increase in aquifer thickness, the RMSE between the modeled and observed heat flux is increased by 45% from the reference simulation (Figure 4b, light square).

Details are in the caption following the image
(a) Measured and modeled surface heat flux along the Kumano transect, presented in the same fashion as in Figure 3. (b) Percent improvement in the misfit between modeled and observed heat flux from simulations with hydrothermal circulation to the reference simulation.

Along the Japan Trench, we examine three transects, labeled A (in the north) through C (in the south) (Figure 2b). For line A, the reference simulation heat flux is ∼50 mW/m2 on the incoming plate and ∼25 mW/m2 on the margin wedge (Figure 5a, dashed line); this is up to 50 mW/m2 lower than the observations on the incoming plate. The model that is most consistent with the observed heat flux values includes an aquifer that thickens to 5 km thick and does not host vigorous fluid circulation landward of the trench (Figure 5b). The thickening of the aquifer mining heat from deeper in the crust generates a surface heat flux anomaly that extends 150 km seaward of the trench; the greatest magnitude of the anomaly (22 mW/m2) occurs immediately seaward of the trench (Figure 5a, bold black line). Landward of the trench, the modeled heat flux in this case is within 10 mW/m2 of that for the reference simulation (Figure 5a). Relative to the reference simulation, this preferred simulation reduces the RMSE between the modeled and observed heat flux by 33% (Figure 5b, star). If vigorous fluid circulation in an aquifer with the same thickness (5 km maximum) continues up to 100 km landward of the trench, the resulting thermal anomaly has a maximum amplitude of 53 mW/m2 (Figure 5a, dark gray line). This is less consistent with the heat flux observation than the preferred simulation; the misfit between the modeled and observed heat flux is only reduced by 10% (Figure 5b, dark square). Relative to the preferred simulation, reducing the maximum aquifer thickness to 0.6 km (i.e., no aquifer thickening) generates a thermal anomaly that extends only 50 km seaward of the trench with a maximum amplitude of 2 mW/m2. In this case with no increase in aquifer thickness, the RMSE between the modeled and observed heat flux is reduced by only 1% from the reference simulation (i.e., 32% less reduction in error than in the preferred simulation; Figure 5b, light square).

Details are in the caption following the image
(a) Measured and modeled surface heat flux for the Japan Trench along line A, presented in the same manner as in Figure 3. (b) Percent improvement in the misfit between modeled and observed heat flux from simulations with hydrothermal circulation to the reference simulation with no hydrothermal circulation.

For line B along the Japan Trench, the preferred simulation includes an aquifer that thickens to 3 km thick with vigorous circulation extending to 100 km landward of the trench (Figure 6a, bold black line). In this case, the modeled heat flux is up to 38 mW/m2 higher than in the reference simulation with no hydrothermal circulation, with the peak in the anomaly immediately seaward of the trench (Figure 6a). The thermal anomaly extends 150 km seaward of the trench (Figure 6a). In this case, the misfit between the modeled and observed heat flux is 15% lower than for the reference simulation (Figure 6b, star). Decreasing the landward extent of vigorous circulation reduces the magnitude of the thermal anomaly immediately seaward of the trench to 13 mW/m2 (Figure 6a, dark gray line). However, there are no heat flux observations from 0–50 km seaward of the trench (i.e., where the modeled heat flux for this simulation and the reference simulation diverge most substantially), in part due to logistical/instrumental limitations and the great water depth at the Japan Trench (Yamano et al., 2014). For the simulation with circulation terminating at the trench, the misfit between the modeled and observed heat flux is 3% lower than for the reference simulation (Figure 6b, dark square). Relative to the preferred simulation, reducing the maximum aquifer thickness to 0.6 km (i.e., no aquifer thickening) generates a thermal anomaly with a maximum amplitude of 13 mW/m2 at the trench (Figure 6a, light gray line). In this case, this misfit between the modeled and observed heat flux is increased by <1% from the reference simulation (i.e., ∼15% less reduction in error than in the preferred simulation) (Figure 6b, light square).

Details are in the caption following the image
(a) Measured and modeled surface heat flux for the Japan Trench along line B, presented in the same manner as in Figure 3. (b) Percent improvement in the misfit between modeled and observed heat flux from simulations with hydrothermal circulation to the reference simulation with no hydrothermal circulation.

For line C along the Japan Trench, the preferred simulation has an aquifer with a maximum thickness of 1.8 km and vigorous circulation stopping at the trench (Figure 7a, bold black line). In this case, the modeled heat flux is 7 mW/m2 higher than in the reference simulation with no hydrothermal circulation; the peak in the anomaly is 8 km seaward of the trench (Figure 7a). The thermal anomaly extends 150 km seaward of the trench (Figure 7a). In this case, the misfit between the modeled and observed heat flux is 5% lower than for the reference simulation (Figure 7b, star). Increasing the landward extent of vigorous circulation to 100 km landward of the trench increases the magnitude of the thermal anomaly relative to the reference simulation immediately seaward of the trench to 30 mW/m2 (Figure 7a, dark gray line). In this case, the misfit between the modeled and observed heat flux is increased by 26% relative to the reference simulation (Figure 7b, dark square). Where there are heat flux observations on the incoming plate, the difference between the modeled heat flux from the preferred simulation and the simulation with circulation extending 100 km landward is <10 mW/m2 (Figure 7a). Relative to the preferred simulation, reducing the maximum aquifer thickness to 0.6 km (i.e., no aquifer thickening) generates extremely small thermal anomalies, within 7 mW/m2 of the reference simulation at all locations (Figure 7a, light gray line).

Details are in the caption following the image
(a) Measured and modeled surface heat flux for the Japan Trench along line C, presented in the same manner as in Figure 3. (b) Percent improvement in the misfit between modeled and observed heat flux from simulations with hydrothermal circulation to the reference simulation with no hydrothermal circulation.

For the simulations at the Japan Trench transects that include anisotropic thermal conductivity in the aquifer (i.e., only high conductivity vertically; no enhanced lateral heat transport), the modeled surface heat flux patterns are very sensitive to the maximum aquifer thickness, but quite insensitive to the landward extent of hydrothermal circulation (Figures S1 and S2 in the supporting information). For lines A and C, the preferred maximum aquifer thicknesses are the same in the simulations with anisotropic and isotropic thermal conductivity in the aquifer (5 km for line A and 1.8 km for line C; Figure S1). For line B, the preferred maximum aquifer thickness in simulations with anisotropic conductivity in the aquifer is 5 km (compared with 3 km for the isotropic simulations; Figure S1). For the cases with anisotropic thermal conductivity in the aquifer, the modeled surface heat flux patterns are fairly insensitive to variations in the landward extent of hydrothermal circulation; there is little difference (<5%) in error reduction between simulations with circulation stopping at the trench and those with circulation extending to 45 km landward of the trench (Figure S1). When isotropic thermal conductivity is used, the magnitude of the modeled surface heat flux anomaly is increased by ∼5–15 mW/m2 for simulations with isotropic thermal conductivity and the same aquifer thickness and landward extent of hydrothermal circulation as the isotropic preferred simulations (Figure S2).

4 Discussion

For the Nankai margin, we focus primarily on the Muroto transect, where there is less scatter in the heat flux observations (compared to the Kumano transect), such that the suite of simulations that produce heat flux consistent with the observations is better constrained. For the Muroto transect, generating a ≥100 mW/m2 anomaly in heat flux near the trench requires vigorous circulation in the aquifer extending >50 km landward of the trench. Increasing the aquifer thickness as the crust approaches the trench is not required in order to generate a surface heat flux anomaly consistent with the observations; however, substantial aquifer thickening is permitted. The simulation that minimizes the misfit between modeled and observed heat flux has no aquifer thickening (i.e., it has a constant 0.6-km-thick aquifer; Figure 3b). Models with a maximum aquifer thickness of 0.6 km produce surface heat flux patterns most consistent with the observations (>20% error reduction relative to the reference simulation), provided the landward extent of vigorous circulation is between 100 and 175 km (Figure 3b). With the landward extent of vigorous circulation between 50 and 100 km, simulations with a maximum aquifer thickness ranging from 0.6 km (i.e., no aquifer thickening) up to 3 km all yield 10–20% error reduction; for these values of the landward extent of vigorous circulation the surface heat flux pattern is not particularly sensitive to the degree of aquifer thickening.

The thermal regime of the Muroto transect is substantially influenced by the presence of continued hydrothermal circulation after subduction (Figure 8a). The isotherms from the reference simulation (Figure S3a) illustrate three key thermal processes: (1) advection of the subducting slab lowering the vertical thermal gradient through the overriding plate, (2) heat generation by friction along the plate interface between the subducting Philippine Sea Plate and the overriding Amurian Plate, and (3) heat advection by mantle wedge flow. The isotherms from the preferred simulation reflect these same three processes, plus the presence of hydrothermal circulation (Figure S3b). In this scenario, vigorous fluid circulation continues 150 km landward of the trench, transporting heat laterally updip and cooling the shallow portion of the plate interface. The temperature difference between the reference and preferred simulations is portrayed in Figure 8a. Here the contours indicate the cooling effects of hydrothermal circulation. The greatest cooling (>90 °C) occurs at ∼150 km landward of the trench (i.e., at the landward end of vigorous fluid circulation in the subducting aquifer and near the peak in frictional heating). Much of the heat imparted to the system by frictional heating on the plate interface is advected seaward, away from this portion of the plate boundary. There is no advection of heat by hydrothermal circulation farther landward of the trench (i.e., >150 km landward); however, hydrothermal cooling farther updip cools the plate interface and upper portions of the subducting slab by >50 °C relative to the reference simulation to ∼350 km landward of the trench (Figure 8a).

Details are in the caption following the image
Effects of hydrothermal circulation on temperatures in the (a) Muroto and (b) Kumano transects. The contours (every 20 °C) are the difference between the reference and preferred simulations, illustrating the amount and distribution cooling by hydrothermal circulation. Black lines are the seafloor, the top of the subducting plate; the dashed line is the boundary between lithospheric continental mantle and asthenospheric continental mantle.

For the Kumano transect, there is more scatter in the heat flux observations at a given distance from the trench than for the Muroto transect (Figure 4a). Much of the scatter in the heat flux observations seaward of the trench in the Kumano transect likely result from substantial spatial variations in sediment thickness above the top of a nearly isothermal basement aquifer (e.g., Spinelli, 2014). For the majority of the simulations for the Kumano transect, the misfit between the modeled and observed heat flux is greater than that for the reference simulation, and the maximum reduction in that misfit relative to the reference simulation is only 4% (Figure 4b). The scatter in the data for the Kumano transect may result from a greater fraction of the data lying farther from the transect, combined with along-strike variability in sediment thickness not accounted for in the 2-D cross section. For the Muroto transect, heat flux observations are tightly clustered along the transect; 269 heat flux observations are within 50 km of the transect; only 35 more are from 50–100 km from the transect (Figure 2a). For the Kumano transect 101 heat flux observations are within 50 km of the transect; another 62 are from 50–100 km distant from the transect (Figure 2a). This scatter hampers our ability to identify preferred simulations for the Kumano transect, as has been noted in previous studies (e.g., Harris et al., 2013); however, using 15-km binned median heat flux values when assessing the ability of each simulation to fit the available data mitigates some of the spatial variability. An additional constraint on the thermal history of the margin along the Kumano transect is provided by the amount of alteration inferred from clay mineral assemblages deep (up to 3 km below seafloor) in the margin wedge (Underwood & Song, 2016; Underwood, 2017). Thermal models that predict the amount and distribution of smectite-to-illite reaction progress most consistent with the observations include hydrothermal circulation in a subducting aquifer extending landward from the trench (Spinelli & Underwood, 2017). Although these models (Spinelli & Underwood, 2017) do not examine the potential role of aquifer thickening, the study provides another indication that fluid circulation likely extends landward of the trench in the subducting aquifer and redistributes heat laterally along much of the Nankai Trough margin.

The thermal effects of hydrothermal circulation are less pronounced for the Kumano transect than for the Muroto transect. The landward extent of hydrothermal circulation for the preferred simulation (35 km landward of the trench) is less than that for the Muroto transect (150 km landward of the trench; Figure 8). Comparing the results from the reference simulations for the Nankai Trough transects (Figure S4a vs. Figure S3a), the Kumano transect is cooler than the Muroto transect due to the steeper slab dip and older subducting lithosphere in the Kumano transect. The isotherms in Figure S4b reflect the presence of hydrothermal circulation in the preferred simulation where vigorous fluid circulation continues to 35 km landward of the trench. Figure 8b shows the temperature difference between the reference and preferred simulations. The cooling effects of hydrothermal circulation are greatest (>50 °C reduction) at the landward end of vigorous fluid circulation in the subducting aquifer (35 km landward of the trench). Hydrothermal circulation reduces temperature along the plate interface by at least 30 °C from the trench to ∼100 km landward of the trench. Compared to the Muroto transect, hydrothermal circulation in the Kumano transect transports less heat laterally, producing a smaller and shallower region of substantial temperature reduction.

For each transect at the Japan Trench, simulations that generate heat flux anomalies most consistent with the observations include substantial aquifer thickening. The preferred simulations for each transect require significant aquifer thickening; 5 km for line A, 3 km for line B, and 1.8 km for line C. Having vigorous circulation in the aquifer extending landward of the trench is not required in order to generate the observed broad (extending >100 km seaward of the trench) low amplitude (≤50 mW/m2) heat flux anomalies. However, lateral heat exchange in the aquifer that extends landward of the trench is permitted. Having vigorous circulation in the aquifer extend 100 km farther landward as in preferred simulation for line B as opposed to terminating at the trench has a limited effect on surface heat flux where the heat flux observations are concentrated. (i.e., an increase <9 mW/m2 relative to the modeled heat flux in the preferred simulations more than 50 km seaward of the trench; Figure 6a); the same is true for lines A and C (i.e., an increase <18 mW/m2 relative to the modeled heat flux in the preferred simulations more than 50 km seaward of the trench). Varying the landward extent of vigorous circulation has the largest effect on surface heat flux immediately seaward of the trench (Figures 5a, 6a, and 7a). Thus, in order to use the surface heat flux pattern to better constrain the landward extent of vigorous circulation for the Japan Trench, additional data from 0–25 km seaward of the trench would be particularly useful. Unfortunately, the water depth in the Japan Trench makes acquiring such data technically challenging. Also, variations in basement topography (line A) are usually accompanied by variations in overlying sediment thickness; this could contribute to noise in the surface heat flux observations from the Japan Trench.

In all three transects across the Japan Trench, aquifer thickening is the dominant characteristic of hydrothermal circulation affecting the thermal state of the system. The isotherms from the reference simulations for lines A–C (Figures S5–S7a) are nearly identical. The >100-million-year-old subducting Pacific Plate produces a very cold subduction zone, where 200 °C on the plate interface is not reached until ∼250 km landward of the trench (Figure S5a). In all three Japan Trench reference simulations, the lateral temperature gradient in the subducting aquifer is much smaller than those for the Nankai Trough transects (Figures S3–S7a); as a result, less heat is laterally redistributed in the Japan Trench regardless how far landward of the trench vigorous fluid circulation occurs (or how high the thermal conductivity is in the subducted basement aquifer). Numerous factors contribute to making the Japan Trench a very cold subduction zone (i.e., old subducting plate, fast convergence, and steep slab dip). Thus, despite the more rapid burial of the subducting crust in the Japan Trench (resulting primarily from the steeper slab dip), relative to that in the Nankai margin, the temperature gradient between the subducted aquifer and the aquifer seaward of the trench is much smaller in the Japan Trench than in the Nankai Trough. For the Japan Trench system, it is not easy (nor effective) for fluid circulation in the aquifer to redistribute a sufficient quantity of heat to generate the high heat flux anomaly near the outer rise; it is more effective for fluid circulation in the crust to mine heat from deeper in the crust locally as a result of aquifer thickening.

The preferred simulation for line A, in which hydrothermal circulation terminates at the trench in an aquifer with a maximum thickness of 5 km, yields a modest cooling effect along the plate interface (Figure S5b). Since hydrothermal circulation is limited to the aquifer seaward of the trench, heat transport is not laterally enhanced across the trench; however, heat is mined from significantly deeper in the oceanic lithosphere due to the thickening of the aquifer prior to subduction. The cooling effects of hydrothermal circulation are greatest (>30 °C reduction) at depths between ∼10 and 50 km (Figure 9a).

Details are in the caption following the image
Effects of hydrothermal circulation on temperatures in the Japan Trench transects: (a) line A, (b) line B, and (c) line C. The contours (every 10 °C) are the difference between the reference and preferred simulations, illustrating the amount and distribution cooling by hydrothermal circulation. Black lines are the seafloor, the top of the subducting plate; the dashed line is the boundary between lithospheric continental mantle and asthenospheric continental mantle.

The thermal effects of aquifer thickening are slightly less marked for line B than for line A. The preferred simulation for line B has an aquifer with a maximum thickness of 3 km; however, hydrothermal circulation continues up to 100 km landward of the trench (Figure S6b). The maximum aquifer thickness is located 45 km landward of the trench. Here, hydrothermal circulation mines heat from deeper within the lithosphere and incorporates it into the aquifer. Since the aquifer is thinner than for line A, this yields a smaller, shallower region of hydrothermal cooling between 10- and 20-km depth (Figure 9b).

The thermal effects of aquifer thickening and hydrothermal circulation between the reference and preferred simulations of line C are <25 °C throughout the entire system (Figure 9c). As in line A, hydrothermal circulation is limited to the aquifer seaward of the trench; therefore, heat transport is not laterally enhanced across the trench. Unlike line A, heat is mined from only the uppermost 1.8 km in the oceanic lithosphere. Cooling along the plate interface is limited to ∼20 °C (Figure 9c). Such a small reduction in temperature is expected to have a small influence on the location and/or rate of progress of thermally controlled processes.

Large vertical offsets along normal faults through the crust of the Pacific Plate subducting at the Japan Trench (e.g., Kimura et al., 2012) could compartmentalize fluid circulation in the oceanic crustal aquifer. Thus, the normal faults may facilitate the mining of heat from deeper in the crust but also disrupt lateral heat redistribution via hydrothermal circulation in the aquifer. Simulations that include enhanced vertical thermal conductivity in the aquifer (but not enhanced lateral thermal conductivity) are most consistent with surface heat flux observations when the maximum aquifer thickness is 1.8–5 km (Figures S1 and S2). Because lateral heat transport is not enhanced by hydrothermal circulation in these scenarios, the modeled surface heat flux patterns are insensitive to the landward extent of vigorous fluid circulation.

The Nankai and Japan Trench margins differ in that contrasting processes are required to generate thermal anomalies that are consistent with the observed surface heat flux patterns. At the Nankai margin, lateral heat exchange between the subducted aquifer and the crust seaward of the trench facilitates the generation of the large amplitude heat flux anomaly focused seaward of the trench. Thickening of the aquifer as the crust approaches the trench may modify the heat flux distribution; however, in isolation, this process is not sufficient to generate the observed thermal anomaly. At the Japan Trench, aquifer thickening as the crust traverses the outer rise can facilitate the generation of the broad low amplitude heat flux anomaly that extends >100 km seaward of the trench. Lateral heat exchange between the subducted aquifer and the crust seaward of the trench may alter the heat flux distribution, but it cannot be the sole process generating the observed thermal anomaly. Comparison of the modeled and observed heat flux patterns permits the possibility that both processes (lateral heat exchange extending landward of the trench and aquifer thickening) are active at both the Nankai and Japan Trench margins. The Kumano transect on the Nankai margin is the one case for which the combined effects of lateral heat exchange extending landward of the trench and aquifer thickening produces a modeled heat flux pattern that is less consistent with the observations than a reference simulation with no heat redistribution in the aquifer.

The difference in heat transport mechanisms could be attributed to the physical attributes of each subduction zone. The distribution of high-permeability faults and/or fractures at depth in the oceanic crust is likely influenced by the plate curvature (Faccenda, 2014). Eventually, fractures will be sealed at depth either by silica precipitation or thermoelastic closure (Lowell et al., 1993). The Philippine Sea Plate at the Nankai Trough is much younger (∼15–20 Ma) than the Pacific Plate at the Japan Trench (∼135 Ma). Based on analysis of receiver functions at borehole seismometers, Kawakatsu et al. 2009 estimate that the Pacific Plate entering the Japan Trench is ∼82 km thick; in contrast, the 25-Ma lithosphere of the Philippine Sea Plate is ∼55 km thick (Kawakatsu et al., 2009). The younger, thinner, and hotter Philippine Sea Plate has a shallower dip (and a smaller degree of curvature) for its subducting slab (20–31°) than the Pacific Plate subducting at the Japan Trench (29–34°; Hayes et al., 2012). The large lateral temperature gradient between the crust seaward and landward of the trench in the hot Nankai Trough subduction zone facilitates long-distance lateral heat exchange in the subducting aquifer (e.g., Rotman & Spinelli, 2013; Spinelli & Wang, 2008). Conversely, the larger degree of curvature for the older, thicker, and colder Pacific Plate results in more extensive aquifer thickening and increases the depth to which vigorous fluid circulation can occur; however, the age and cold nature of the incoming plate offsets the lateral temperature gradient generated by the steeper slab dip (Rotman & Spinelli, 2013). Significant aquifer thickening (extensive plate deformation) is not likely in the Philippine Sea Plate of the Nankai Trough, given its shallow curvature and therefore lower magnitude of plate-bending stresses.

5 Conclusions

The Japan Trench and Nankai Trough subduction zones exhibit different physical characteristics with regard to age, slab dip, and temperature. This makes them excellent candidates for determining the thermal effects of aquifer thickening prior to subduction as well as the effects of hydrothermal circulation. Our models indicate a greater sensitivity to the thermal effects of aquifer thickening prior to subduction for the Japan Trench than for the Nankai Trough. This can be attributed to the large degree of curvature of the subducting Pacific plate creating a thicker aquifer that mines heat from deeper in the basement and incorporates it into the aquifer. The hotter Nankai Trough subduction zone (with a shallow dip of the Philippine Sea plate) has a larger temperature gradient between the subducted slab and the plate seaward of the trench; thus, fluid circulation in the oceanic crustal aquifer transports heat laterally seaward. Future work should be done applying this technique to other subduction zones with physical characteristics within the range of these two extreme examples to further understand the implications of aquifer thickening on surface heat flux.

Acknowledgments

We thank Patrick Fulton, Rob Harris, and an anonymous reviewer for constructive comments that improved this study. Heat flux data presented in the study are available from previous publications (Watanabe et al., 1970; Yamano et al., 1984; Kinoshita & Yamano, 1986; Yamano et al., 1992; Yamano et al., 2003; Fulton et al., 2013; Harris et al., 2013; Yamano et al., 2014). Output from model runs for this study are presented in the figures in this manuscript. Spinelli was supported by National Science Foundation Grant OCE 1551587.