Large submarine landslides in the Japan Trench: A new scenario for additional tsunami generation



[1] We describe in detail possible large submarine landslides, several tens of kilometers in length and width, on the trench landward slope of the Japan Trench on the basis of high-resolution topographic surveys and detailed seafloor observations. These slides stopped at the toe of the trench slope. After initial movement of the toe along a basal decollement or thrust of the trench landward slope wedge during an earthquake, the basal frictional condition(s) might change drastically from static to dynamic, thus reducing the frictional strength. As a result, rapid submarine landsliding push downward on the toe, generating large horizontal displacements for tsunamis. This hypothesis should explain suitably the relation between large displacement of the thrust fault and tsunami generation by the 2011 Tohoku earthquake as well as tsunami generation by the 1896 Tohoku earthquake.

1. Introduction

[2] Tohoku, northeast Japan, experienced a great earthquake (moment magnitude; Mw 9.0, tsunami magnitude; Mt 9.1) (Figure 1) [e.g., Ide et al., 2011]. Seismic and tsunami inversion analyses have shown that tsunami waves with a maximum run-up height of more than 40 m around Miyako (39°48′N) [Mori et al., 2011] were generated just after the main-shock by topographic changes on the seafloor in the toe region of the Japan Trench slope off Sendai (37°30′N–38°30′N) as shown byYokota et al. [2011]. These inversion analyses [Maeda et al., 2011; Yokota et al., 2011], bathymetric surveys [Fujiwara et al., 2011] and ocean-bottom pressure observations [Ito et al., 2011] indicate that the toe region slipped about 40–80 m along the thrust.

Figure 1.

(a) Topographic map of the southern part of the Japan Trench (500 m mesh J-EGG bathymetric data) and locations of cross sections. Vertical exaggeration is 6 times. Inset shows the study area and the epicenter of the 2011 Tohoku earthquake (star). (b) Trench cross sections showing the typical landward slope morphology. Vertical exaggeration is 5 times. The lateral distance line of origin is the maximum depth along the trench axis. The cross sections follow the seismic lines ofTsuru et al. [2002]. Dashed red lines are at the water depth of 6000 m. (c) Stacked profiles of the landward lower slope of the Japan Trench. Colors indicate the latitude ranges along the trench axis. Vertical exaggeration is 5 times. The origin is at the maximum depth of each E-W cross section along the trench axis. The dashed yellow line shows the mean lower slope angle of 5°.

[3] However, seismic inversion analyses have indicated that the thrust fault slipped about 20 m at the hypocenter [e.g., Ide et al., 2011; Yokota et al., 2011]. If the thrust fault rapidly deformed the seafloor, as suggested by Ide et al. [2011], “the basic theory of tsunami genesis” induced by seafloor deformation by rupture propagation would predict the generation of tsunamis all along the axis of the Japan Trench as shown by the scenario of Tanioka and Satake [1996a].

[4] However, the results of the tsunami inversion analyses do not seem to be consistent with this scenario. Maeda et al. [2011]suggested on the basis of data from two ocean bottom pressure gauge stations off Kamaishi (39°12′N), 45–75 km offshore, which detected the first maximum tsunami waveform, that the tsunami source was a narrow region in the deeper part of the Japan Trench slope off Sendai. In fact, the tsunami run-up heights in the northern Tohoku area cannot be explained by a single source off Sendai [Satake et al., 2011]. The other source must have generated a tsunami after the first wave and main earthquake shock. These lines of information imply that the second tsunami wave from the north was not generated directly by earthquake-induced thrust.

[5] This region is approximately the source area of the tsunami generated by the 1896 Sanriku tsunami earthquake, that is, the toe of the trench slope between 38°15′N and 40°10′N [Tanioka and Satake, 1996b]. Abe [1979] showed that in spite of Ms (surface seismic wave magnitude) being small as 7.2, Mt was as large as 8.6. In order to explain the diversity between Ms and Mt, the 1896 Sanriku earthquake has been attributed to slow rupture along the fault [Kanamori, 1972], or to submarine landslides [Kanamori and Kikuchi, 1993], or to additional rapid uplift of a sediment wedge [Tanioka and Seno, 2001]. What is correct?

2. Topographic Analyses and Deep-Sea Camera Observations

[6] The Japan Trench slope can be divided into an upper slope, a mid slope terrace, and a lower slope (Figures 1a and 1b). The average slope angle of the upper and lower slopes is 5°, but that of the mid slope terrace is only 2° [von Huene and Lallemand, 1990]. Some upward convex parts of the upper and lower slopes attain more than ∼10°, particularly in the region from 39°10′N to 40°30′N with Sanriku Escarpment as a fault cliff (Figure 1c). The upper slope includes many large, upward convex, horizontally arcuate topographic features suggesting submarine landsliding of sedimentary rock masses with widths and lengths of several kilometers (Figure 2) [Sasaki, 2004]. These submarine landslides have been explained by tectonic erosion as shown by von Huene and Lallemand [1990].

Figure 2.

(left) Arcuate topography and (right) structural map. The bathymetric contour interval is 20 m. Filled circles A–D indicates YKDT Dive points. The green dotted line shows an area of sliding between spurs. Scarps and slope failures and channels have developed on the lowermost part of the upper slope. Photographs of the seafloor: (a) YKDT Dive 94; (b–d) YKDT Dive 97.

[7] The upper slope is characterized by many normal faults of mostly N-S strike [Tsuru et al., 2002; von Huene and Lallemand, 1990]. Fresh N-S open fractures with gray-colored-rock surfaces were observed with the deep-sea camera system on 8 June 2011 duringYokosuka Deep Tow (YKDT) Dive 94 in the normal fault region of one submarine landslide, strongly supporting very fresh fracturing (Figure 2a). On the lower foot slope, we detected pressure ridges on the seabed in high-resolution bathymetric maps (Figure 2) and found fresh N-S open fractures and sharp scars on the ridges during YKDT Dive 97 on 11 June 2011 (Figures 2b and 2c). These ridges were probably formed when the submarine landsliding pushed the seabed into a sediment wedge (Figure 2). We also observed a recent collapse, consisting of large blocks and gravels accumulated at the foot of the pressure ridges (Figure 2d). These features indicate that the sliding is still occurring, and also that sliding might have been triggered by the 2011 Tohoku earthquake.

3. Tsunami Genesis by Submarine Landsliding in the Subduction Zone

[8] According to seismic inversion analyses, slip with dynamic overshoot may have occurred at the toe of the trench slope [Ide et al., 2011]. This result from rapid elastic response and requires small dynamic basal friction of the wedge for overshooting. After the initial thrusting along the basal decollement or thrust fault, this dynamic overshooting is followed by smooth slip on the basal plane under dynamic frictional conditions, which are proposed by Seno [2002] that “a transition from stable sliding to nearly zero friction, due to velocity weakening, occurs as a result of the growth of zones of elevated fluid pressure, which may make a fairly rapid seismic slip possible following the breakage of asperities at normal seismogenic depths”. High fluid pressure in the basal decollement zone of the Japan Trench slope was already suggested by seismic images [von Huene and Culotta, 1989]. This horizontal displacement of a sediment wedge at the toe of the trench slope is equivalent to the fundamental mechanism for generation of a large tsunami in the Japan Trench as shown by Tanioka and Seno [2001].

[9] On the basis of geologic architectures in the Japan Trench, we consider that the toe of the slope moves by the combined effect of seismic slip and submarine landsliding (Figure 3). After initial movement of the toe along a basal decollement or thrust during an earthquake, the basal frictional condition(s) might change drastically from static to dynamic, thus reducing the frictional force. As a result, rapid submarine landsliding might push downward on the toe, generating large thrust displacements with particularly horizontal into the eventual tsunami. Such landslides do explain easily why the topographic displacement becomes larger toward the toe of the trench slope.

Figure 3.

Schematic diagram of tsunamigenic slip at the toe of the trench slope. The submarine landslide was potentially under the instable conditions before the earthquake, and it was moved with rupture propagation along the thrust. The slip at the toe is accelerated by submarine landsliding. The many normal faults at the head of the submarine landslide originated as fault bifurcations.

4. Concluding Remarks

[10] It has been believed that tsunamis are generated only by seafloor topographic change caused only by active faulting. However, the Japanese tsunami warning system does not include the tsunami excitation scenario by submarine landsliding. In fact, in 1979, a tsunami 2–3 m in height struck Nice, France, unaccompanied by any seismic signals. This silent tsunami was considered to be generated by submarine landsliding near the Nice harbor [Dan et al., 2007]. Thus, all data pertaining to tsunami generation mechanisms as well as topographic changes in survey data from before to after the 2011 Tohoku earthquake should be carefully examined to improve our understanding of tsunami generation.


[11] We acknowledge the leadership and assistance of Katsunori Fujikura (Japan Agency for Marine-Earth Science and Technology), the Chief Scientist for cruise YK11-E04 of RVYokosuka, and the captain, crew, and operation teams of the Yokosuka Deep Towsystem. This study was supported in part by a Grant-in-Aid from the Ministry of Education, Culture, Sports, Science and Technology of Japan to the first author (20340139).

[12] The Editor thanks Andrea Festa and an anonymous reviewer for their assistance in evaluating this paper.