Laboratory measurements quantifying elastic properties of accretionary wedge sediments: Implications for slip to the trench during the 2011 M_w 9.0 Tohoku-Oki earthquake Open Access

Rock properties are a major controlling factor in frictional behavior of faults and earthquake rupture dynamics (Scholz, 2002). The shear modulus, or modulus of rigidity, is of particular interest because it describes a rock’s resistance to shear deformation and influences earthquake rupture velocity and fault slip. Low shear modulus materials in the shallow regions of subduction zones have been cited as the cause of the shallow aseismic zone (e.g., Byrne et al., 1988) and of greater than expected tsunami wave heights experienced during tsunami earthquakes (e.g., Kanamori, 1972; Fukao, 1979; Pelayo and Wiens, 1992; Bilek and Lay, 1999a; Tanioka and Seno, 2001). Earthquake data have been used to estimate shear modulus variations with depth (Dziewonski and Anderson, 1981; Bilek and Lay, 1998, 1999a, 1999b, 2000), but these earthquake-scale studies do not estimate shear modulus in the upper ∼5 km of a subduction zone. Therefore, shallow shear modulus variations remain poorly constrained and are typically ignored in megathrust studies in general.

Numerical models of dynamic rupture processes by Lotto et al. (2017) have explored the effect of accretionary prisms composed of compliant sedimentary material on subduction zone earthquakes. They found that a compliant prism can slow rupture velocity, increase shallow slip, and increase tsunami height relative to models without a compliant prism. The degree to which velocity, slip, and wave height are altered depends on the size of the prism, contrast in shear modulus between the prism and surrounding material, and basal friction. The effect of a prism on subduction zone earthquakes is complex, so a better understanding of the elastic properties of prisms at convergent margins is needed to fully characterize seismic and tsunami hazard.

We have performed laboratory measurements of ultrasonic wave speed and density on samples of drill core obtained during Integrated Ocean Drilling Program (IODP) Expedition 343, known as the Japan Trench Fast Drilling Project (JFAST). These measurements allow us to compute the dynamic elastic moduli of the frontal prism of the Japan Trench. Dynamic elastic moduli, rather than static, are most relevant to this study because we are interested in the processes occurring during the rapid propagation of fault rupture and slip, which occurs at speeds close to the shear wave velocity. In this paper, we show that the Japan Trench frontal prism is very compliant relative to standard crustal model values, and we explore global variations in shallow subduction zone elastic moduli. In addition, we used the laboratory data with wide-angle refraction data from the Japan Trench (Miura et al., 2001, 2005) to create a 2D shear modulus model of the Japan Trench that is applicable to compliant prisms globally and can be used to explore their role in coseismic shallow slip and tsunami generation.

At the Japan Trench subduction zone off the northeastern coast of Japan, the Pacific plate subducts beneath Honshu Island at a shallow angle of 5° to 15° (Fig. 1) (von Huene and Langseth, 1982; Apel et al., 2006; Ide et al., 2011; Satake et al., 2013). The Pacific plate is composed of basaltic rocks overlain by a chert unit and a sequence of pelagic clay and hemipelagic mudstone, all cut by high-angle normal faults (Arthur and Adelseck, 1980; Chester et al., 2013; Nakamura et al., 2014). A relatively small frontal accretionary prism has formed landward of the trench. The prism is acoustically chaotic and composed of Neogene sediments, specifically hemipelagic mudstones and underthrust terrigenous mudstones and pelagic clays (Chester et al., 2013; Nakamura et al., 2014; Kirkpatrick et al., 2015; Moore et al., 2015). The backstop of the prism is formed by a sequence of Cretaceous rocks that are separated from Neogene slope cover sediments by an erosional surface (von Huene and Culotta, 1989; von Huene and Lallemand, 1990; von Huene et al., 1994). This sequence was sampled at Deep Sea Drilling Program (DSDP) Site 439 (Murauchi and Ludwig, 1980).

The Japan Trench megathrust fault hosted the Tohoku-Oki (M_W 9.0) earthquake on 11 March 2011. There is evidence that the rupture reached the seafloor at the trench with up to 60 m of displacement at shallow depths and produced a large tsunami (Ammon et al., 2011; Fujii et al., 2011; Ide et al., 2011; Ito et al., 2011; Koper et al., 2011; Lay et al., 2011; Kodaira et al., 2012; Sun et al., 2017). Previous to this earthquake, it was generally assumed that the velocity-strengthening frictional properties and low strength of sediments in shallow parts of subduction zones would cause megathrust slip to dissipate rather than grow as rupture traveled through this energy-absorbing, compliant, and velocity-strengthening material. Thus the large shallow slip of the Tohoku-Oki earthquake raised many questions about potentially anomalous frictional and elastic properties of the Japan Trench subduction zone. Since that event, high-velocity friction experiments have shown that rupture can propagate through the shallow subduction zone relatively easily due to a decrease in strength and fracture energy at high slip velocities (Faulkner et al., 2011; Ujiie et al., 2013).

One year after the Tohoku-Oki earthquake, JFAST drilling was undertaken. The main goal of JFAST was to understand the physical mechanisms and dynamics of the large slip earthquake (Mori et al., 2012). Three boreholes were drilled in the frontal prism of the Japan Trench subduction zone at Site C0019, located ∼200 km east of the Oshika Peninsula (Fig. 1). Drilled to depths of ∼845–855 m below the seafloor (mbsf), the boreholes were used to (1) collect geophysical well logs (natural gamma ray and resistivity), (2) retrieve core samples, and (3) install a temperature observatory (Expedition 343/343T Scientists, 2013). Based on changes in log signature and bedding orientation observed in borehole logging data from Hole C0019B, the plate-boundary fault was interpreted to be located ∼820 mbsf.

Coring in Hole C0019E targeted the probable fault location identified in the log data. Seven lithologic units were identified in the recovered core: (1) olive-gray mudstone (176.5–185.2 mbsf); (2) brown and bluish gray mudstone (648.0–659.7 mbsf); (3) gray mudstone (688.5–820.1 mbsf); (4) sheared clay (821.5–822.5 mbsf); (5) brown mudstone (824–832.9 mbsf); (6) pelagic clay (832.9–833.5 mbsf); and (7) chert (833.5–836.8 mbsf). Unit 4, sheared clay, is interpreted to represent the plate-boundary décollement zone (Expedition 343/343T Scientists, 2013).

Laboratory high-pressure experiments to measure ultrasonic wave velocity were performed on seven samples from the JFAST site C0019E borehole (Table 1). Cylindrical specimens with diameters of 25.4 mm were extracted from the JFAST cores with the axis of the specimen oriented perpendicular to the drill core axis. All laboratory samples were prepared by hand using precision knives and sandpaper due to the delicate nature of the sample material. Six of the samples are of the Unit 3 gray mudstone located above the plate boundary fault. The seventh sample comes from the underthrust brown mudstone of lithologic unit 5.

Following pulse-transmission techniques adapted from Birch (1960, 1961) and Christensen (1985) velocity measurements were performed on the cylindrical specimens under elevated confining and pore-pressure conditions (Table 1).

All samples except 7R1 were measured at the University of Wisconsin–Madison Rock Physics Laboratory in the Department of Geoscience. Sample 7R1 was measured at Kochi University. At the University of Wisconsin, samples were loaded into a standard triaxial vessel (Fig. 2) with sintered metal filters at both ends to prevent extrusion of the sample material into the pore pressure ports when pressure is applied. Six 500 kHz central frequency piezoelectric transducers embedded in the endcaps of the vessel (three in each endcap) were used as source-receiver pairs. These transducers generate the P-wave and two mutually perpendicular S-wave vibration directions used in the wave propagation experiments.

During our experiments, confining pressure, axial stress, and pore-fluid pressure were independently controlled using three syringe pumps; salt water (salinity of 35 ppt) was used as a saturating fluid. The axial and confining pressures were simultaneously increased to 2 MPa and 1.8 MPa, respectively. The 0.2 MPa difference between axial stress and confining pressure was maintained throughout the experiment because the design of the experimental apparatus requires axial stress to be greater than confining pressure to prevent leakage between the sample liner and endcap. The specimen was then saturated by creating a pressure differential across the sample. One side of the sample was left open to atmospheric pressure, and a 1 MPa pore pressure was applied at the other side. Once water began flowing out the open end of the sample, that side was connected to the syringe pump so the saturating fluid flowed into both ends of the sample. Pore space gas should be minimal because core samples were stored in conditions that limited drying; thus, samples were still wet when placed in the vessel. Pore-fluid pressure of 1 MPa was maintained through the experiments so that any remaining pore-space gas was kept in solution, ensuring 100% saturation. The specimen was allowed to saturate for ≥24 h before the first ultrasonic velocity measurement was made. We then simultaneously increased the confining and axial pressure to obtain an effective pressure of 2 MPa and allowed the sample to equilibrate. Ultrasonic wave-speed measurements were conducted at regular intervals between 1 MPa and 70 MPa. Following each pressure step, the sample was allowed to equilibrate, as monitored by continuous measurements of sample length and pore volume change, before an ultrasonic measurement was made. Waveforms were recorded during both increasing and decreasing pressure paths. P- and S-wave speeds were determined from time-of-flight measurements and the measured sample length.

At Kochi University, sample 7R1 was measured using a similar procedure, but pore pressure was held constant at 0.5 kPa, and confining pressure was only increased to 18 MPa. Additionally only one source-receiver pair of identical S-wave polarized 500 kHz transducers was used, similar to the method described in Gettemy and Tobin (2003). This transducer arrangement yields only one waveform instead of the three individual waveforms recorded at Wisconsin. This increases the uncertainty in determining the S-wave arrival. Preliminary analysis of the data indicated that the S-wave measured at Kochi was noticeably different from the S-waves measured at Wisconsin and resulted in an unrealistic Poisson’s ratio that decreased with pressure. For this reason, we only used the P-wave measurement from this sample.

Initial picking of the wave arrivals had a high level of uncertainty, particularly in the S wave at low pressures. We have decreased that uncertainty by cross-correlating waveforms acquired at low pressures with those recorded at higher pressures. This improved the wave arrival pick at low pressures where poor sample-transducer coupling leads to greater uncertainty. Overall there is less than 1% uncertainty in picking the P-wave arrival and a 2% uncertainty in the S-wave arrival (Figs. S1 and S2¹). This uncertainty was based on repeated waveform measurements that were acquired at each pressure step and our ability to repeatedly pick the same arrival on different instances of the waveform. The uncertainty decreases with increasing pressure and improved coupling. Additional uncertainty in the velocity arises from non-parallel core faces and associated uncertainties in initial length measurements. The core faces can usually be smoothed and made parallel so there is less than 0.5 mm variability leading to 2% uncertainty in the length of the sample. The total uncertainty in the velocity measurement is less than 4%.

Final porosity and density were determined through direct measurement of sample volume and wet and dry mass. These measurements, combined with the axial displacement, were used to estimate porosity and density changes during the experiment. The total sample volume was determined using both linear measurements using a vernier caliper and volumetric displacement methods. Because the samples are fragile and must be prepared by hand, the sample diameter is not uniform. This results in a relatively high uncertainty of 6% in the density measurements and 7% in the porosity measurements.

Velocity was measured over a cycle as effective pressure was increased from 1 to 70 MPa and then decreased back to 1 MPa (Figs. 3 and S3 [^{footnote 1}]). The pressure curves indicate that the sediments are very compliant. The samples experience permanent axial strains of up to 14%. The shallowest sample, 4R2, is the most compliant. There is also hysteresis in the velocity curves: at 15 MPa, there is an 8%–19% difference in the P-wave velocity and a 13%–24% change in the S-wave velocity between the loading and unloading values. The velocity values increase with increasing effective pressure for all samples measured.

The measured wave speeds are low (v_p <3.2 km/s, v_s <1.6 km/s) and similar in both the unit 5 underthrust brown mudstone specimen and the unit 3 gray mudstone samples. The laboratory velocities are consistent with seismic-reflection models that indicate the velocity should range between 1.5–3.0 km/s in the frontal prism sediments in the Japan Trench (Murauchi and Ludwig, 1980; Miura et al., 2001, 2005; Nakamura et al., 2014). They also agree with ultrasonic wave-speed measurements of prism sediments from other subduction zones (Carson and Bruns, 1980; Tobin et al., 1994, 1995; Tobin and Moore, 1997; Gettemy and Tobin, 2003; Hashimoto et al., 2010; Raimbourg et al., 2011; Schumann et al., 2014). For the purposes of this study, we are mainly interested in the average elastic properties of the prism. Because of this, and the limited number of samples available for this study, fine-scale variations are only briefly discussed in the following paragraphs.

In the unit 3 gray mudstone samples examined here, an overall trend of increasing velocity with sample depth is apparent. However, sample 8R2 deviates from that trend because it has a slower velocity than samples located at shallower depths. The lower velocity could be due to a higher porosity, since velocity typically has an inverse relationship with porosity (Nafe and Drake, 1957; Kowallis et al., 1984; Han et al., 1986; Klimentos, 1991; Hashimoto et al., 2010). A porosity of 48% was measured for sample 8R2; this is the highest porosity we measured for the unit 3 samples. Comparison of our porosity measurements to porosities measured during the JFAST expedition on the same cores (Expedition 343/343T Scientists, 2013) shows that the porosities differ by less than 10% for samples collected at depths greater than 713 mbsf. However, the porosities measured for our two shallowest samples (4R2 and 6R1) are 15% less than the on-expedition measurements, indicating that our measurements of these two samples may be in error. The expected porosity range for 4R2 and 6R1, based on shipboard measurements, is between 48% and 53% (Fig. 4). Because the difference between the initial and final length measurements for theses samples doesn’t correlate with the recorded displacement during the experiment, it is likely that the samples experienced some initial consolidation as the vessel endcaps were brought into contact with the sample before we began recording sample displacement. These samples are therefore over-consolidated relative to the other samples measured. This would explain the higher velocity and lower porosity of 4R2 and 6R1.

The unit 5 brown mudstone sample (19R2) is the deepest sample examined in this study, but it has a slower velocity than the samples directly above it (15R1 and 16R1). An examination of discrete velocity measurements acquired during the JFAST expedition on unsaturated, unconfined samples, shows a general trend of increasing velocity with depth in unit 3 but a decrease in velocity in the unit 5 underthrust sediments relative to the basal wedge sediments of unit 3 (Fig. 4A) (Expedition 343/343T Scientists, 2013). Porosity is variable but generally decreases with depth; so the porosity variations do not explain the decrease in velocity. Based on X-ray diffraction (XRD) analysis by Kameda et al. (2015), unit 5 has greater clay content than unit 3. The smectite content in unit 3 remains relatively constant with an average of 13 wt% of the bulk samples, whereas in unit 5, the average smectite content is 18 wt% of the bulk sample (Kameda et al., 2015). Velocity has been shown to have an inverse relationship with clay content; so the lower velocity in unit 5 could be due to the higher clay content (Tosaya and Nur, 1982; Kowallis et al., 1984; Castagna et al., 1985; Han et al., 1986; Klimentos, 1991).

The dependence of velocity on clay content is also seen in data from DSDP Site 436 (Fig. 4B) (Carson and Bruns, 1980). X-ray diffraction data from Kameda et al. (2015) show a smectite-rich horizon at both sites (gray shaded regions in Fig. 4). Immediately below the top of this horizon, the velocity decreases relative to the velocities above even though porosity is decreasing. Although the smectite content is greater at Site 436 than at C0019E and there are several age reversals with depth at Site C0019E due to folding and faulting (Chester et al., 2013; Moore et al., 2015), the smectite content below the clay-rich horizon is greater than in the sediments above the horizon at both sites. Although the sediments at Site C0019E are not in the same stratigraphic sequence as those from Site 436, similar dependence of velocity on clay content is observed.

The frontal prism sediments from the Japan Trench are characterized by low bulk modulus, low shear modulus, and high Poisson’s ratio (Figs. 5 and S4 [^{footnote 1}]). The moduli vary systematically with effective pressure, as reflected in increased P- and S-wave speeds with effective pressure. At an effective pressure of 5 MPa on the loading curve, which is comparable to in situ conditions, the bulk modulus ranges from 4.7 to 9.9 GPa, shear modulus ranges from 1.1 to 2.3 GPa, Poisson’s ratio ranges from 0.33 to 0.41, and v_p/v_s ratio ranges from 1.99 to 2.58. This is consistent with other evidence that indicates the shallowest parts of subduction zones have low shear moduli relative to crustal rocks (shear modulus ∼30 GPa), which numerous workers have suggested causes an updip limit to the seismogenic zone (e.g., Fukao, 1979; Bray and Karig, 1985; Byrne et al., 1988; Satake, 1994; Lay and Bilek, 2007).

The large magnitude of shallow coseismic slip to the trench during the 2011 (M_w 9.0) Tohoku earthquake was unexpected due to the predicted low modulus of rigidity and velocity strengthening properties of sediments near the trench. At other subduction zones, it has been suggested that the occurrence of shallow coseismic slip is due to the presence of unusually competent and strong sediments which allow the accumulation and release of elastic energy (Gulick et al., 2011). To test this hypothesis, we examined ultrasonic velocity measurements that have been collected on shallow accretionary prism sediments retrieved from other subduction zone margins by IODP, DSDP, and Ocean Drilling Program (ODP) projects (Tobin et al., 1994, 1995; Tobin and Moore, 1997; Gettemy and Tobin, 2003; Hashimoto et al., 2010; Raimbourg et al., 2011; Schumann et al., 2014). We compared our computed elastic properties to published data from other accretionary prisms to address whether a difference in the elastic properties could account for the occurrence of coseismic slip to the trench during the Tohoku earthquake.

Many published studies did not report S-wave speed or density for accretionary prism materials. To estimate the S-wave speed for the prism materials of other convergent margins, we used the data from this study, along with Gettemy and Tobin’s (2003) data from the Costa Rica margin, Schumann et al.’s (2014) data from the Nankai Trough, and sonic log data from the NanTroSEIZE project Site C0002P (Webb, 2017) to derive a v_p-v_s relationship (Fig. 6). Over the range of values, a linear fit provides a good approximation of the data. Because the data set is small and dominated by samples from the Nankai Trough, we compared this relationship to the v_p-v_s relationship derived by Brocher (2005) using measurements of a variety of common lithologies (not including marine sediments) and the v_s-depth relationship derived by Hamilton (1976) for marine and terrestrial sediments. The average uncertainty between these estimates of v_s is 0.09 km/s.

Where density was not reported, IODP, DSDP, and ODP databases of shipboard measurements and lithologic descriptions were used to obtain the nearest neighbor (always less than 1.5 m) density measurement within the same lithologic unit. To calculate dynamic elastic moduli, we used the velocity recorded at the estimated in situ pressure for each sample. The results of this comparison are shown in Figure 7.

The Japan Trench data fit in the overall depth trend of the other subduction zone moduli. The Japan Trench has lower Poisson’s ratio than most of the samples acquired at other subduction zones, but without more data at depths greater than 600 mbsf, it cannot be determined if the Japan Trench or Nankai Trough data are more representative of global accretionary prisms. While it is difficult to identify distinct rigidity-depth trends in the data, there are some differences among the prisms. The majority of data presented in Figure 7 come from the Nankai Trough, where multiple boreholes have been drilled at different locations in the accretionary prism as part of the NanTroSEIZE project. These data show variations in dynamic elastic moduli related to the setting from which each sample was taken within the accretionary prism volume. The variations appear to be controlled by differences in composition, grain size, and history. For example, at the same depth below the seafloor, samples from the prism toe are less compliant than samples from the upper accretionary prism (Fig. S5 [^{footnote 1}]) and contain 5–10 wt% less smectite clay (Guo and Underwood, 2012). The lower clay content may explain the greater elastic moduli of the prism toe samples. Underthrust samples have smectite content similar to the upper accretionary prism and appear to lie on a similar rigidity-depth trend. If the apparent rigidity-depth trends of the prism toe and upper accretionary prism were projected to greater depths, both types of sediment would be more rigid than the forearc basin sediments. Clay content in the forearc basin is similar to that of the prism toe (Guo and Underwood, 2012), but the forearc basin sediments have not experienced the tectonic stresses and deformation that sediments from other regions of the prism have experienced. These stresses and deformation are expected to cause greater compaction leading to less compliant prism sediments, as proposed by Jarrard (1997).

Most of the samples from the Oregon prism and Costa Rica margin have elastic properties similar to those of the Nankai accretionary prism sediments, but the Barbados Ridge samples are more compliant. Total clay content in the Barbados Ridge samples ranges from 50 to 80 wt%, whereas at Nankai, it ranges from 30 to 70 wt% (Underwood and Deng, 1997; Guo and Underwood, 2012). The greater clay content may explain the more compliant nature of the Barbados Ridge samples.

Overall, variations in the elastic properties among the different subduction zones are small. Without a more complete data set, we cannot define a substantial difference between the elastic properties of the Japan Trench and the other margins examined here. This contradicts the hypothesis that anomalously rigid elastic properties of the Japan Trench permitted anomalously shallow coseismic slip (although such behavior likely does require a large-magnitude earthquake in any setting).

Laboratory data are not available to constrain the elastic properties of the subduction zone at depths greater than a few kilometers below the seafloor. Although samples have been subjected to greater stresses in the laboratory (up to 70 MPa) to simulate greater depths, geologic modification related to diagenesis is not duplicated in such experiments. The shear modulus at greater depths in the crust has been estimated by inverting seismic normal mode data and body-wave travel times (Dziewonski and Anderson, 1981), and shear modulus specifically along subduction zone megathrust faults has been estimated based on rupture duration and source depth (Bilek and Lay, 1999b, 2000). Figure 8 shows that different data sets used to estimate the shear moduli at a variety of scales (laboratory, log, and earthquake) are not inconsistent with each other, but if the megathust rigidity-depth relationship of Geist and Bilek (2001) or the preliminary reference earth model (PREM) (Dziewonski and Anderson, 1981) was projected to the surface, the shear modulus of the prism would be overestimated by up to 20 GPa.

To better characterize the shallow shear modulus variations in the Japan Trench, we combined our laboratory results with data from a wide-angle refraction study (Miura et al., 2001, 2005) to create a cross section of shear modulus variation within the Japan Trench. The refraction data were acquired on the MY 102 seismic line, located north of Site C0019 (Fig. 1). Density was estimated based on laboratory data and an existing density model developed by Miura et al. (2005). In the Neogene slope cover sequence, Cretaceous marginal wedge, upper crust, lower crust, and incoming sediment layers, the Brocher (2005) regression fit was used to estimate S-wave velocity from refracted P-wave velocities. The Brocher fit could also have been used to estimate the S-wave velocity in the prism, but we used the linear v_p-v_s relationship discussed earlier because it provides a better fit to the laboratory data (Fig. 6). In ocean layers 2 and 3, S-wave velocity was estimated using the Brocher (2005) mafic line equation. The Brocher (2005) v_p-v_s relationships result in v_p/v_s ratios similar to those found in other studies for similar materials (e.g., Christensen, 1989; Hole et al., 1991; Holbrook et al., 1992; Zhao et al., 1992, 2010; Planke and Cambray, 1998; Barclay et al., 2001; Nakajima et al., 2001). For the mantle, we assumed a v_p/v_s ratio of 1.77 based on previous studies (Birch, 1964; Jordan, 1976; Duffy and Anderson, 1989; Zhao et al., 1992; Nakajima et al., 2001).

The resulting shear modulus cross section (Fig. 9) shows that the frontal prism, Neogene cover sequence, and incoming sediments all have very low shear moduli on the order of 1 GPa. There are discrete boundaries at the top of the Cretaceous marginal wedge and ocean layer 2 where the shear modulus abruptly increases to 25 GPa. A depth transect through the prism indicates the PREM relationship provides a good estimate of the shear modulus except in the prism and Neogene cover sequence (Fig. 9B), suggesting that a different rigidity-depth relationship is appropriate for poorly consolidated sediments.

We lack measurements of samples from the megathrust fault itself, and the fault is too narrow to be resolved by the wide-angle refraction data; therefore, we cannot provide any constraints on shear modulus within the megathrust. Assuming that the properties of the sub-prism megathrust are similar to the material properties in the hanging and foot walls, then the Geist and Bilek (2001) relationship, which was derived from Central America earthquake data, would overestimate the shear modulus of the fault. Shear moduli computed using IODP expedition 348 sonic log from depths of 2–3 km in the Nankai Trough accretionary prism (Webb, 2017) are well described by the Geist and Bilek (2001) relationship (Fig. 8), suggesting that it may be more applicable to subduction zones with a larger accretionary prism. However, unless the fault is very thick, it is not expected that the shear modulus of the fault itself would have a significant effect on slip propagation.

The frontal prism of the Japan Trench is a volumetrically small component of the subduction zone and is therefore commonly ignored in numerical simulations of earthquake rupture. Most models assume a constant shear modulus on the order of 30 GPa and a Poisson’s ratio of 0.25 throughout the subduction zone (e.g., Ide et al., 1993; Liu and Rice, 2005; Loveless and Meade, 2011; Sun et al., 2017). This is comparable to the PREM shear modulus and is suitable for most crustal rocks. However, tsunamigenic earthquakes have been linked to the presence of low shear moduli material in the shallow subduction zone, and quantitative models indicate a low shear modulus is required to match model results to observations of tsunami amplitude (Kanamori, 1972; Fukao, 1979; Okal, 1988; Pelayo and Wiens, 1992; Kanamori and Kikuchi, 1993; Satake, 1994, 1995; Tanioka and Satake, 1996; Heinrich et al., 1998; Satake and Tanioka, 1999). In most tsunamigenic earthquake models, the shear modulus is only reduced to 10–20 GPa, and the prism is not included in the models. Although waves propagating in low-shear modulus materials are attenuated more rapidly than in higher-shear modulus materials, Hooke’s Law shows that the stress generated by the wave will result in greater strain. Incorporation of a compliant prism in earthquake models may therefore lead to an amplification of simulated fault slip and tsunami wave height. This effect might explain where and how the largest tsunamis are generated.

Numerical models of Lotto et al. (2017) examined the effect of variations in prism size, shear modulus, and basal friction on shallow slip magnitude and tsunami wave height. In these models, it is apparent that rupture velocity slows dramatically in the prism to a velocity close to the prism S-wave velocity. The presence of a small (less than 10-km-wide) compliant prism results in local enhancement of shallow slip that is amplified, if the shear modulus is on the order of 2 GPa or if the basal frictional properties are velocity weakening. The presence of a large prism (greater than 20 km wide) increases shallow slip, if the frictional properties are velocity-weakening but decreases shallow slip, if the frictional properties are velocity-strengthening. This effect is amplified if the prism has very low shear modulus. Small, velocity-strengthening prisms have relatively little effect on tsunami wave height, but if the prism is velocity-weakening or larger in size, the effect on tsunami wave height is greater. Thus, although the effect of a prism is clearly complex, the presence of a compliant prism does influence fault slip and tsunami generation.

For a small prism similar to the frontal prism of the Japan Trench, where S-wave velocity is ∼1 km/s and sub-prism friction is neutral to velocity-strengthening (Ikari, 2015; Ikari et al., 2015), the results of Lotto et al. (2017) predict a 7–21 m increase in slip at the trench relative to a model where there is no prism. The tsunami wave height would change by 0.5 m or less. Dynamic rupture simulations of the 2011 M_w 9.0 Tohoku-Oki earthquake predicted ∼30 m of slip at the toe of the trench, if velocity-strengthening friction extends 30 km down dip from the trench, or 40 m of slip, if velocity-strengthening friction only extends to 15 km (Kozdon and Dunham, 2013). These simulations did not account for the effects of a compliant prism and did not match the displacements observed nearest the trench. Incorporating a compliant frontal prism, consistent with our measurements of in situ material properties, could produce the additional 7–20 m of displacement needed to match the 50–60 m near trench slip (Fujii et al., 2011; Ide et al., 2011; Ito et al., 2011; Lay et al., 2011; Kodaira et al., 2012) that was observed during the Tohoku-Oki earthquake.

Laboratory ultrasonic wave-speed measurement of seven samples from the JFAST core helps quantify the elastic moduli of the shallow subduction zone. Comparison to data from other subduction zone margins shows that the elastic properties of the Japan Trench frontal prism are similar to those of frontal prisms at other margins. For all samples we examined, the bulk modulus, shear modulus, and Poisson’s ratio range from 3 to 10 GPa, 0.2–3 GPa, and 0.35–0.5, respectively, at estimated in situ pressures. This indicates that anomalous material properties, relative to other subduction zones around the world, are not responsible for the unexpected large magnitude slip to the trench experienced during the 2011 Tohoku earthquake. Furthermore, by logical extension, it is likely that Tohoku-like slip to the trench behavior is also not anomalous, although it requires a large-magnitude earthquake.

Shallow coseismic slip extending all the way to the seafloor during large-magnitude earthquakes has been observed at the Japan Trench (Fujii et al., 2011; Ide et al., 2011; Ito et al., 2011; Lay et al., 2011; Kodaira et al., 2012), Sunda Trench (Henstock et al., 2006; Gulick et al., 2011), and Peru-Chile Trench (Yue et al., 2014). Evidence of past coseismic slip to the trench has also been found at the Nankai Trough (Sakaguchi et al., 2011) and Cascadia margin (Satake et al., 2003). Because the compliant elastic properties of subduction zone prisms have been linked to increased shallow slip and greater tsunami amplitudes, incorporation of shallow shear modulus variations is necessary for accurate simulations of earthquake wave propagation and displacement through the shallow subduction zone. Laboratory, borehole log, and earthquake estimates of shear modulus from global subduction zones are not inconsistent with one another, but rigidity-depth relationships based solely on earthquake-scale estimates overestimate the shear modulus of the frontal prism. We anticipate that incorporation of the shallow shear modulus variations in the Japan Trench, presented in this study, into simulations of the 2011 Tohoku-Oki earthquake will produce a better fit to observations of shallow, near-trench slip. Incorporating compliant prisms into numerical simulations of other subduction zones is similarly expected to provide better estimates of the seismic hazard posed by different margins through an improved understanding of the magnitude of coseismic slip to the trench and the mechanics of tsunami earthquakes.

Laboratory data used in this study were acquired at University of Wisconsin–Madison and Kochi University. These data and the elastic property models are available from the author upon request. We thank the science party of Expedition 343/343T and the staff of the Drilling Vessel Chikyu for their dedication and assistance in collecting the data and samples used in this study. Thanks to Gabriel Lotto, Eric Dunham, Ludmila Adam, Brett Carpenter, and an anonymous reviewer for their comments on earlier versions of this manuscript. A U.S. Scientific Support Program Post-Expedition award to Jeppson funded this work.