How large peak ground acceleration by large earthquakes could generate turbidity currents along the slope of northern Japan Trench

Deep-sea turbidite has been used to determine the history of occurrence of large earthquakes. Surface-sediment remobilization is a mechanism of the generation of earthquake-induced turbidity currents. However, the detailed mechanism of surface-sediment remobilization caused by earthquake ground shaking is unclear. To understand how high peak ground acceleration (PGA) caused by a large earthquake can remobilize surface sediments, we determined the age of a surface-sediment core recovered from the mid-slope terrace (MST) of the inner slope of the Japan Trench in northern Sanriku to determine turbidites generated by large historical earthquakes and calculate the PGAs of these earthquakes using an empirical attenuation relation commonly used in Japan. Small offsets in radiocarbon ages and excess 210Pb activities between turbidite and hemipelagic muds suggest that the turbidites in the core resulted from surface-sediment remobilization. 137Cs and excess 210Pb chronologies indicate that the three uppermost turbidites in the core are correlated with three large historical earthquakes, namely the 1968 common era (CE) Tokachi-oki, the 1933 CE Showa–Sanriku, and the 1896 CE Meiji–Sanriku earthquakes. Calculation of PGAs for large historical earthquakes along the northern Japan Trench indicates that a PGA of > 0.6 g is necessary for turbidite deposition in the MST basin. This threshold is larger than that reported for central Sanriku and may vary spatially. Moreover, turbidites in the MST deposits are more frequent in the northern Japan Trench than in the central Japan Trench, suggesting that the occurrence of three types of large M8-class earthquakes in the northern Japan Trench might have contributed to the frequent occurrence of large PGAs.


Introduction
Deep-sea turbidite has been used to determine the history of occurrence of large earthquakes (e.g., Adams 1990;Goldfinger et al. 2007Goldfinger et al. , 2012Patton et al. 2015;Ikehara et al. 2016). Large earthquakes are one of the main mechanisms of the initiation of turbidity currents, although other mechanisms have been proposed, such as storm surges, large storm waves, floods (hyperpycnal flows), rapid sediment loading, submarine groundwater discharge, volcanic eruption, and bolide impacts (e.g., Goldfinger et al. 2012;Pickering and Hiscott 2015). While earthquake-induced submarine landslides and transformation from landslides to turbidity currents through subaqueous debris flows, as exemplified by the west Atlantic (Heezen and Ewing 1952;Heezen et al. 1954;Piper et al. 1999), are a major mechanism for generating earthquake-induced turbidites, another mechanism has recently been proposed. This involves the surfacesediment resuspension and remobilization caused by large ground shaking during large earthquakes (Moernaut et al. 2017). The obtained remobilization depths in the Chilean lakes range between 0.6 and 8.5 cm, with an average of 5.3 cm (Moernaut et al. 2017). The occurrence of turbidity currents of surface-sediment remobilization origin has been studied for several recent and historical large earthquakes (Moernaut et al. 2014(Moernaut et al. , 2017McHugh et al. 2016;Molenaar et al. 2019Molenaar et al. , 2021Ikehara et al. 2020Ikehara et al. , 2021Schwestermann et al. 2021). However, there is little knowledge on how large earthquake ground shaking causes surface-sediment remobilization. Although several papers have discussed the minimum peak ground acceleration (PGA) caused by large earthquakes that is needed to generate slope instabilities and turbidite depositions (e.g., Lee et al. 1999;Lykousis et al. 2002;Strasser et al. 2007;Noda et al. 2008;Dan et al. 2009;Pouderoux et al. 2014;Usami et al. 2018), the responses of slope sediments to earthquake ground shaking are spatially different, reflecting variations in topography (e.g., slope angle and roughness) and sedimentology (e.g., sediment thickness and grain size and composition). Therefore, it is crucial to gather examples of surface-sediment remobilization generated by historical earthquakes and understand the minimum PGA caused by large earthquakes that are necessary for initiation of turbidity currents from surface-sediment remobilization.
The mid-slope terrace (MST) is a deep-sea terrace with a water depth of 4000-6000 m; it is formed along a series of faults at the front of the backstop of the subducting Pacific Plate along the central to northern Japan Trench (von Huene et al. 1980;Tsuru et al. 2002). Sediments supplied by gravity flows from the landward slopes are trapped at this flat terrace (von Huene and Culotta 1989;Tsuru et al. 2002;Usami et al. 2018;Ikehara et al. 2020). Gravity flows by surface-sediment remobilization is a mechanism of earthquake-induced downslope sediment transport along the MST (McHugh et al. 2016;Molenaar et al. 2019;Ikehara et al. 2020Ikehara et al. , 2021. Turbidites generated by recent and historical earthquakes were reported from the central MST (McHugh et al. 2016;Usami et al. 2018). Therefore, the MST along the Japan Trench is an ideal area for examining the threshold of PGA caused by large historical earthquakes that are needed to create turbidity currents via surface-sediment remobilization.
In this paper, we examine a surface-sediment core collected from the northern MST, where recent and historical earthquakes have occurred in different patterns from the central MST, to determine turbidites formed by large historical earthquakes. We discuss the minimum PGA needed to generate surface-sediment remobilization along the slope.

The mid-slope terrace along the northern Japan Trench and surface-sediment core
At the Japan Trench, the Pacific Plate subducts northwestward the Okhotsk Plate at a rate of 8.0-8.6 cm/year (DeMets et al. 2010). The continuous fault activity along the Japan Trench produces a backstop interface and the MST has been formed along the faults at the backstop (von Huene et al. 1980;Tsuru et al. 2002). The sediments transported by gravity flows from the inner (landward) slope are trapped in a series of isolated depressions of the MST with little lateral migration of sediments (von Huene and Culotta 1989;Tsuru et al. 2002;Usami et al. 2018;Ikehara et al. 2020). The lack of submarine canyon connecting the shelf and the MST off the Sanriku slope suggests that storm-and flood-related turbidity currents can hardly reach the MST and earthquakes are the most likely mechanism of the generation of the turbidity currents on this slope. The sediments at the MST are generally diatomaceous (Mann and Müller 1980;Usami et al. 2018;Ikehara et al. 2020Ikehara et al. , 2021, reflecting high diatom productivity in the surface water (Saino et al. 1998). High average sedimentation rates are observed at the MST (100-110 cm/ky), and the MST deposits comprise the alternation of bioturbated hemipelagic muds and muddy turbidites (Usami et al. 2018;Ikehara et al. 2020). Several large (M > 7.6) and recent earthquakes have occurred around the central-northern Japan Trench (Tajima et al. 2013;Satake 2015;Toda 2016;Lay 2018; Headquarters for Earthquake Research Promotion 2019). These were the 2011 CE off the Pacific coast of Tohoku earthquake (Tohokuoki; M 9.0, Mw 9.1), 1994 CE Sanriku-oki earthquake (M 7.6,Mw 7.8 We obtained core YK14-E01 PL08 (location: N 40°08.0264′, E 143°57.9590′; water depth: 4203 m; core length: 95 cm) from a small basin at the MST off Kuji, northern Sanriku, using a gravity corer during the YK14-E01 cruise of R/V Yokosuka ( Fig. 1; http:// www. jamst ec. go. jp/ datad oi/ doi/ 10. 17596/ 00016 54. html). Because the gravity corer was used as a trigger of a piston corer, free fall was not used, and an undisturbed sample of the sediment-water interface was obtained. The coring site was at the same location as the piston core (YK14-E01 PC08), which was used by Ikehara et al. (2020).

Determination of event deposits
The core was cut vertically, and the sedimentary structures, sediment color, grain size, and grain composition on the cut surface of one-half were logged visually. X-radiographs of 1 cm thick slab samples were obtained at the Geological Survey of Japan, AIST, to observe their detailed sedimentary structures and radiodensity. Magnetic properties, such as mean magnetic susceptibility (K), magnetic lineation (L), and magnetic foliation (F), were measured using continuously collected standard paleomagnetic plastic cube samples (7 cm 3 ) and a KLY-4 magnetic susceptibility meter (Agico Co.) at JAMSTEC. A total of 78 discrete samples were analyzed: 41 from core PL08 and 37 from the uppermost Sect. (85.5 cm in length) of core PC08. The mean magnetic susceptibility and two magnetic fabric parameters (lineation and foliation) were obtained from three eigenvectors, namely, K 1 , K 2 , and K 3 , which are the maximum-, intermediate-, and minimum-susceptibility axes, respectively (Hrouda 1982;Tarling and Hrouda 1993). K = (K 1 + K 2 + K 3 ) / 3 (mean susceptibility; Nagata 1961). L = K 1 / K 2 (magnetic lineation; Balsley and Buddington 1960). F = K 2 / K 3 (magnetic foliation; Stacey et al. 1960). The horizons of muddy turbidites were determined using a combination of sedimentary and magnetic signatures. Small offsets in radiocarbon ages between the turbidite muds and the underlying or overlying hemipelagic muds are useful for studying the surface-sediment remobilization of turbidite origin (Okutsu et al. 2019;Ikehara et al. 2020). The radiocarbon ages of five turbidite mud samples and five hemipelagic mud samples were determined using bulk organic carbon in sediments to determine the age offset between the two sample groups. The bulk sediment samples used for measurement, sized 5-10 cm 3 , were washed with acid to remove carbonates prior to measurement. The stable carbon isotope ratios (δ 13 C) of the same samples for radiocarbon dating were determined using an isotope-ratio mass spectrometer. The radiocarbon dates and stable carbon isotopes were measured in the accelerator mass spectrometer radiocarbon facility at Beta Analytics Co. Ltd.

Age determination
Radioactivity for 137 Cs, 210 Pb, and 214 Pb in 20 sediment samples of 8-18 g in wet weights was measured at 1-2 cm intervals by Environchron Inc. at Eckerd College Marine Science Department (Florida). The samples were ground to fine powder, weighed, and sealed in a 4 cm diameter counting container. We used a Mirion (Canberra) high-purity germanium coaxial planar photon detector linked to a multichannel analyzer (Mirion (Canberra) DSA 1000), and the measurement time was ~ 24-50 h. Data were corrected for emission probability at the measured energy, counting time, and sample mass and then converted to activity (disintegrations per gram, dpm/g) using the International Atomic Energy Association (IAEA) organic standard IAEA-477 for calibration (Kitto 1991;Larson et al. 2018). Excess 210 Pb was defined as the difference between the measured activities (dpm/g) of supported and unsupported 210 Pb and supported 210 Pb, which was determined by averaging the activity of its parent radioisotopes, 214 Pb and 214 Bi, assuming the uranium series had reached secular equilibrium (Kato et al. 2003). The excess 210 Pb-derived age model was calculated using a constant initial concentration model assuming constant excess 210 Pb concentration upon sedimentation (Cundy and Croudace 1995; Kanai 2000). The 137 Cs activity profile (peak and first occurrence of 137 Cs activity) was used to determine the start of atmospheric nuclear bomb experiments, which were detected in 1955 CE, and the peak of 137 Cs fall-out in 1963 CE in Japan (Hirose et al. 1987(Hirose et al. , 2008Igarashi et al. 1996). For the correlation of turbidites in cores PL08 and PC08 (Ikehara et al. 2020), which were obtained from the same location, excess 210 Pb and 137 Cs activities were also measured for nine samples of hemipelagic muds from the upper part of core PC08.
The radiocarbon ages of bulk organic carbon in hemipelagic muds are important for sediment age determination in the environment below the calcium carbonate compensation depth. However, previous study (e.g., Ikehara et al. 2020) has reported that there is a larger uncertainty in the depositional ages estimated from the radiocarbon ages than those estimated from 137 Cs and excess 210 Pb chronology. Therefore, herein, we used the radiocarbon ages of bulk organic carbon only for discussing the age offset between hemipelagic and turbidite muds and not for sediment age determination.

Estimation of PGA
We estimated the maximum acceleration caused by historical earthquakes on the inner slope of the coring site. Six large historical earthquakes that occurred around the coring sites were selected for calculation: the 2011 CE Tohoku-oki, 1994CE Sanriku-oki, 1968CE Tokachioki, 1933CE Showa-Sanriku, 1931 CE Sanriku-oki and 1896 CE Meiji-Sanriku earthquakes. Several methods have been proposed for PGA calculation (e.g., Fukushima and Tanaka 1990;Zhao et al. 1997;Si and Midorikawa 1999). We used an empirical attenuation relation that is commonly used in Japan and derived from records of ground motion on land (Si and Midorikawa 1999) and is the same as that proposed by Usami et al. (2018), who examined the PGA for turbidite depositions at the central MST, because we would like to discuss the spatial variability of PGA along the slope of the Japan Trench.
where where x is the shortest distance from the fault plane (fault distance), D is the focal depth (average depth of the fault plane), and d corresponds to the type of earthquake (crustal: 0.00; interplate: 0.01; intraplate: 0.22). We applied d = 0.01 in the calculations because the studied earthquakes were all interplate earthquakes except for the 1933 CE Showa-Sanriku earthquake, which was an outer-rise earthquake within the subducting plate. However, we used the same d value for the 1933 CE Showa-Sanriku earthquake in this calculation. We calculated PGA at an assumed source point of sediments at the landward slope near our coring site, namely the GeoB21818 coring site (N 40°14.790, E 143°48.918′), which was on a slope at a 3139 m water depth (the location shown in Fig. 1B), where sedimentation gaps caused by earthquake-induced surface-sediment remobilization were found in a surface-sediment core sample (Molenaar et al. 2019). Focal depths of 24, 10, 10, and 7 km were used for the 2011 CE Tohoku-oki (Yoshida et al. 2011), 1994CE Sanriku-oki (Nagai et al. 2001, 1933 CE Showa-Sanriku (Kanamori 1971), and 1896 CE Meiji-Sanriku earthquakes (Satake et al. 2017), respectively. For the 1968 CE Tokachi-oki and 1931 CE Sanriku-oki earthquakes, a focal depth of 10 km, which is the same depth as that of the 1994 CE Sanriku-oki earthquake, was set assuming all ruptured the same asperity (Nagai et al. 2001). Strong ground motion stops increasing with Mw for earthquakes with Mw > 8.3 (called magnitude saturation) when using fault distance for calculation (Si et al. 2016). For the 2011 CE Tohoku-oki earthquake, we used Mw 8.4 for calculation. Because this attenuation formula was developed from on land records and high-frequency seismic waveguides more efficiently in the subducting Pacific Plate than in the overriding (landward) plate (Furumura and Kennett 2005), the actual PGAs in the deep-sea close to the Pacific Plate may be stronger.

Sediment lithology and turbidite recognition
A total of six coarse-grained layers (T-L1-T-L6) consisting of a basal coarse-grained part and an overlying muddy part were identified in bioturbated diatomaceous muds ( Fig. 2 and Table 1). Each coarse-grained layer is 1-3 cm in thickness and characterized by a sharp base and fining-upward grading (Fig. 2). Parallel-or crosslaminations are observed in some basal parts. A homogeneous structure without bioturbation is a characteristic feature of the upper part. These features suggest that the basal coarse-grained layers and overlying homogeneous muds are fine-grained turbidites (Stow and Shanmugam 1980). Turbidite muds and hemipelagic muds can be distinguished by the following characteristics: sedimentary structure, radiodensity (X-ray transparency), sediment color, and magnetic parameters ( Fig. 2; Additional file 1: Table S1). Turbidite muds are originally homogeneous in structure and olive black in color and show an increase in radiodensity, mean magnetic susceptibility (> ~ 0.0004 SI), and magnetic foliation (> ~ 1.02), although they are occasionally disturbed by post-depositional burrowing. By contrast, hemipelagic muds are completely bioturbated, have lower radiodensities, are gray-grayish olive-olive black in color, and have lower mean magnetic susceptibility (< ~ 0.0004 SI) and magnetic foliation (< ~ 1.04). Magnetic lineation in turbidite muds is slightly higher (> ~ 1.01) than that in hemipelagic muds (< ~ 1.01). High magnetic foliation in turbidite muds, which has been reported in muddy turbidites or homogenites in Original edge-to-face clay contact patterns of the flocs were preserved in the aggregates, whereas the flocs were broken after deposition, and a randomly oriented structure was formed in hemipelagic muds (Kase et al. 2016). Such differences in clay fabric between turbidite and hemipelagic muds may result in higher magnetic foliation in turbidite muds than in hemipelagic muds. Although the origin of high magnetic foliation remains unclear, magnetic foliation is a useful proxy for differentiating turbidite and hemipelagic muds.

Surface-sediment remobilization as revealed by offset of radiocarbon ages and excess 210 Pb activities between turbidite and hemipelagic muds
The offsets in radiocarbon ages between the turbidite muds and the underlying or overlying hemipelagic muds in core PL08 are small (− 50-120 years; Table 2) and close to those in core PC08 (− 40-330 years except for two turbidites; Ikehara et al. 2020), where most turbidites in core PC08 are the results of surface-sediment remobilization (Ikehara et al. 2020). The offsets in excess 210 Pb activities between the turbidite and hemipelagic muds in core PL08 are also small (Fig. 2). Age offset between the 2011 CE Tohoku-oki earthquake turbidite and the underlying hemipelagic muds is small in a central Japan Trench basin (Bao et al. 2018). The 2011 CE turbidites in the nearby central Japan Trench basins at ~ N 38° contain 137 Cs and excess 210 Pb (Oguri et al. 2013;Ikehara et al. 2016) and fresh organic matter (Schwestermann et al. 2021). Thus, the small age offset between the turbidite and hemipelagic muds suggests little mixture of old organic matter from deep subsurface sediments in the turbidite muds and is considered a proxy for surface-sediment remobilization (Okutsu et al. 2019;Ikehara et al. 2020). The small offsets in excess 210 Pb activities suggest that origin of turbidite muds is the young surface sediments containing similar activities to the hemipelagic muds. High activities of excess 210 Pb were identified in the event deposits formed by the 2011 CE Tohoku-oki earthquake and its associated tsunami .  The stable carbon isotope ratios (δ 13 C) of bulk organic carbon range from − 19.95 to − 20.77 ‰ (− 19.95 to − 20.77 ‰ for the turbidite muds and − 19.97 to − 20.57 ‰ for the hemipelagic muds; Table 2). The lack of a significant difference in the δ 13 C values between the turbidite and hemipelagic muds in core PL08 suggests a deep-marine source of the turbidite muds. Therefore, offshore marine surface sediments are the most likely source of turbidite muds.
Although massive tsunamis are possible trigger mechanisms to transport shallow marine and coastal sediments to deep sea (Arai et al. 2013;Ikehara et al. 2021), the δ 13 C values suggest that the tsunamis are not likely the mechanism for the formation of the turbidites in this core.

Excess 210 Pb and 137 Cs age model and correlation of turbidites with historical earthquakes
A peak of 137 Cs activity (0.61 dpm/g) occurs at a depth of 7 cm just above the uppermost turbidite (T-L1; Fig. 2 and Table 3). 137 Cs activity is not detected below 10 cm from the core top. The absence of 137 Cs below T-L1 and its stable detection above T-L1 suggest that T-L1 was formed near the peak fall-out of 137 Cs around the Japanese islands in 1963 CE (Hirose et al. 1987(Hirose et al. , 2008Igarashi et al. 1996). Excess 210 Pb activities are low in the uppermost 3 cm (~ 20 dpm/g) and peak at a depth of 7 cm (73.55 dpm/g); they decrease toward a depth of 26 cm (2.98 dpm/g) and then fall below the detection limit (< ~ 2 dpm/g) at depths deeper than 38 cm ( Fig. 2 and Table 3). Assuming the core top 2014 CE and a linear relationship between excess 210 Pb activities and the turbidite-free core depths of the hemipelagic muds, which is established by removing the turbidites from the core, the depositional ages of each analyzed horizon are calculated from the constant initial accumulation model (Table 3).
Based on the estimated depositional ages from excess 210 Pb and 137 Cs activities, the three uppermost turbidites can be correlated with three historical earthquakes of the northern Japan Trench. The uppermost turbidite (T-L1) is just below the peak of 137 Cs activity, suggesting that it occurred around the 1960s CE, although the excess 210 Pb-based age model implies that T-L1 was deposited between 1948 and 1934 CE. Pb-based age estimation may afford relatively constant and low excess 210 Pb activities in the uppermost 3 cm of core PL08. Usami et al. (2018) reported that turbidites formed by the 1896 CE Meiji-Sanriku earthquake were just above the detection limit of excess 210 Pb in two cores off central Sanriku: NT13-19 PC08 (N 39°16.4448, E 143°56.6842′) and NT13-19 PC10 (N 39°07.2353′, E 143°54.1586′). According to Molenaar et al. (2019), the surface-sediment gap on the northern Japan Trench slope (core GeoB21818) resulted from the 1896 CE Meiji-Sanriku earthquake and was near the detection limit of excess 210 Pb. Thus, the event deposits from the 1896 CE Meiji-Sanriku earthquake are widely distributed more than 100 km along the northern-central Japan Trench slope and can be recognized using the profile of excess 210 Pb activity.

PGA threshold for initiation of earthquake-induced surface-sediment remobilization along the northern Japan Trench landward slope
Strong earthquake ground shaking is essential for earthquake-induced surface-sediment remobilization (Moernaut et al. 2014(Moernaut et al. , 2017Usami et al. 2018;Molenaar et al. 2019Molenaar et al. , 2021Ikehara et al. 2021). Table 4 summarizes the PGAs at the assumed sediment source at the landward slope of our coring site caused by the studied historical earthquakes determined using the observed or estimated fault parameters. The PGA needed to form earthquakeinduced turbidites at our coring site is approximately 0.6 g (Fig. 3). No turbidite is recognized by historical earthquakes of < 0.6 g at the sediment source. Therefore, based on the examples of three historical earthquakes, the turbidites occurred in the studied MST basin are the most likely record of past large earthquakes of > ~ 0.6 g at the source slope. Based on the profile of excess 210 Pb activity of the surface sediments of core GeoB21818, which was collected updip of our coring site, Molenaar et al. (2019) stated that a few centimeters to 12 cm sediment remobilization occurred due to three large recent and historical earthquakes (2011 CE Tohoku-oki, 1968 CE Tokachioki, and 1896 CE Sanriku-oki earthquakes). However, we found no evidence of turbidite deposition by the 2011 CE Tohoku-oki earthquake, which has a PGA of 0.35 g, in core PL08. Instead, we found turbidite by the 1933 CE Showa-Sanriku earthquake, with a PGA of 0.63 g. Molenaar et al. (2019) proposed an earthquake-magnitude control rather than PGA control on surface-sediment remobilization. However, our turbidite results suggest clear PGA control (> 0.6 g) for turbidite deposition in the northern MST sediments. Distance from the rupture area to the coring site influences the PGA magnitude and thus turbidite deposition. The small-scale and lowconcentration nature of the turbidity current caused by  Fig. 1B the 2011 CE Tohoku-oki earthquake might be a possible reason for the absence of the corresponding turbidites in the MST sediments despite the occurrence of smallscale (2 cm thick) surface-sediment remobilization on nearby slopes (Molenaar et al. 2019). Ikehara et al. (2012) reported a similar absence of turbidites after small-scale turbidity current events caused by the 1997 CE Izu-tohooki earthquake swarm in western Sagami Bay, which was observed at the deep-sea observatory (Iwase et al. 1997).
Another possibility is the short-distance transport of resuspended and remobilized sediments due to smallscale turbidity currents. Ikehara et al. (2021) discussed such short-distance transport for homogeneous muddy event deposits formed by the 2011 CE Tohoku-oki earthquake on the unconfined slopes offshore of southern Sanriku and Sendai. There is no signature of surfacesediment remobilization by the 1933 CE Showa-Sanriku earthquake in the sediments of a nearby slope (Molenaar et al. 2019), which may suggest uneven resuspension of surface sediments on the slope by earthquake ground shaking reflecting spatial variabilities in sediment availability, geotechnical properties and compaction state of surface sediments and/or multiple turbidite sources in the MST sediments. According to Molenaar et al. (2021), the slopes of Chilean lakes have patchy, centimeter-scale gaps in sediments. The six turbidites in the gravity core PL08 are likely correlated with the turbidites in the piston core PC08 (Ikehara et al. 2020) acquired in the same location, based on sediment lithology, the detection limit of excess 210 Pb, and profiles of magnetic parameters ( Fig. 4; Additional file 1: Table S1). The detection limit of excess 210 Pb is below T-L3 in core PL08 and below T-3 in core PC08.
Although the values of mean magnetic susceptibility differ between cores PL08 and PC08, the three uppermost turbidites in core PC08 (T1-T3) are correlated with those in core PL08 (T-L1-T-L3). A 26.1 cm thick amalgamated turbidite in core PL08 (T-L4-T-L5) and another turbidite (T-L6) might be correlated with the T4 and T5 turbidites in core PC08, according to their stratigraphic position. The thick nature of turbidite muds is common in both T-L4 and T4. The correlation of amalgamated turbidite (T-L4-T-L5) with single-bed turbidite (T4) may suggest multiple sources of a turbidite event and large heterogeneity of turbidite deposition within a small distance. As suggested by these turbidite correlations between the two cores, disturbance and/or sediment loss at the core top during piston corer penetration and core recovery is/are small in core PC08. Ikehara et al. (2020) identified frequent turbidite deposition (26 turbidites/2300 years) in core PC08. Hence, the studied small basin at the northern MST might record past large earthquakes with PGAs of > ~ 0.6 g near the coring site if the geotechnical and compaction states of surface sediments over the past ~ 100 years have remained the same in the past. The turbidite frequency in core PC08 is higher than those in two cores at the central MST at ~ N 39° reported by Usami et al. (2018) (~ 12 turbidites/4000 years), although the threshold of PGA (0.4-0.5 g) needed to form earthquake-induced turbidites at the central MST is smaller than that at the northern MST (> ~ 0.6 g). The origin of this spatial difference in the PGA threshold for earthquake-induced turbidite deposition along the landward slope of the MST is unclear. The grain size and composition of MST deposits do not change along the MST remarkably. Although the differences in thickness and sedimentation rate of the slope sediments are possible factors, no detailed information about such differences has been obtained. The existence of three types of large (> Mw ~ 8) earthquakes in the northern Japan Trench likely contributes to the higher frequency of turbidites in the northern MST (~ N 40°) than in the central MST (~ N 39°). The first type is the interplate earthquakes in the northernmost part of the Japan Trench, such as the 1968 CE Tokachi-oki earthquake. Nagai et al. (2001) suggested that the rupture of a single asperity can produce an M7-class earthquake, such as the 1994 CE Sanrikuoki earthquake, whereas the rupture of multiple asperities can cause M8-class earthquakes, such as the 1968 CE earthquake. The second type is tsunami earthquakes near the trench axis, such as the 1896 CE Meiji-Sanriku earthquake. The third is outer-rise earthquakes caused by the normal faulting of the subducting Pacific Plate, such as the 1933 CE Showa-Sanriku earthquake. There are no clear criteria for distinguishing these types of large earthquakes in marine turbidite records. Deep-sea turbidite deposition caused by a large earthquake occurred all around the rupture area (e.g., the 1993 CE Hokkaido-Nansei-oki earthquake (Ikehara and Usami 2007) and the 2004 CE Sumatra earthquake (Patton et al. 2015)); the spatial distribution of event deposits by each earthquake may be useful for differentiating and characterizing these types. Another possibility of the higher frequency of turbidites in the northern MST is the increase in the PGA threshold in the last few hundred years. If the PGA threshold in the past was smaller than expected, large earthquakes of < 0.6 g could form turbidites in the MST basin. However, no evidence of this possibility exists. It is important to accumulate the results of analysis of more sediment cores along the entire MST for understanding the spatiotemporal variability of the PGA threshold and its controlling factors.

Conclusions
To understand how a large PGA is necessary for earthquake-induced surface-sediment remobilization and turbidite deposition, we determined the ages of turbidites in a surface-sediment core PL08 recovered from the northern part of the MST along the inner slope of the Japan Trench. Small offsets in radiocarbon ages and excess 210 Pb activities between turbidite and hemipelagic muds suggest that surface-sediment remobilization may be a major mechanism of turbidite deposition. 137 Cs and excess 210 Pb chronologies indicate that the three uppermost turbidites in core PL08 are correlated with three large historical earthquakes along the northern Japan Trench: the 1968 CE Tokachi-oki (Mw 8.2), 1933 CE Showa-Sanriku (Mw 8.4), and 1896 CE Meiji-Sanriku (Mw ~ 8.1) earthquakes. We calculate PGAs for large recent and historical earthquakes along the northern Japan Trench using an empirical attenuation relation commonly used in Japan and derived from records of ground motion on land and find that a PGA of ~ 0.6 g is necessary for turbidite deposition in this MST basin. In the MST basins off central Sanriku, turbidites can be deposited by much smaller PGAs (0.4-0.5 g; Usami et al. 2018). Therefore, the PGA threshold for turbidite deposition in the MST basins is spatially variable. This variability should be more thoroughly examined when considering paleoseismic history. The occurrence of three types of large M8-class earthquakes in the northern Japan Trench may contribute to a higher frequency of turbidites in the northern part of the MST (~ N 40°) than in the central part (~ N 39°).

CE
Common era MST Mid-slope terrace PGA Peak ground acceleration IAEA International Atomic Energy Association