Recurring and triggered slow-slip events near the trench at the Nankai Trough subduction megathrust

See allHide authors and affiliations

Science  16 Jun 2017:
Vol. 356, Issue 6343, pp. 1157-1160
DOI: 10.1126/science.aan3120

Silently taking up the slack

Megathrust earthquakes occur when locked subduction zone faults suddenly slip, unleashing shaking and causing tsunamis. However, seismically silent slow earthquakes also relieve slip on these dangerous faults. Araki et al. present data from ocean boreholes with which they analyze eight slow-slip events near the Nankai trench off the coast of Japan. These events accommodated up to half of the plate convergence over 6 years. The events appear to occur regularly, which has a long-term impact on hazard assessment for the region.

Science, this issue p. 1157


The discovery of slow earthquakes has revolutionized the field of earthquake seismology. Defining the locations of these events and the conditions that favor their occurrence provides important insights into the slip behavior of tectonic faults. We report on a family of recurring slow-slip events (SSEs) on the plate interface immediately seaward of repeated historical moment magnitude (Mw) 8 earthquake rupture areas offshore of Japan. The SSEs continue for days to several weeks, include both spontaneous and triggered slip, recur every 8 to 15 months, and are accompanied by swarms of low-frequency tremors. We can explain the SSEs with 1 to 4 centimeters of slip along the megathrust, centered 25 to 35 kilometers (km) from the trench (4 to 10 km depth). The SSEs accommodate 30 to 55% of the plate motion, indicating frequent release of accumulated strain near the trench.

Over the past several years, geodetic and seismic observations have revealed a continuous spectrum of slow earthquake behaviors, in which failure occurs on time scales ranging from seconds to years, fundamentally changing the way fault slip behavior is understood (17). These phenomena span from normal earthquakes to low-frequency tremors (LFTs) and very-low-frequency earthquakes (VLFEs), characterized by a higher abundance of low-frequency energy in radiated seismic waves than for a typical earthquake (1, 8, 9), and further to slow-slip events (SSEs), in which slip occurs over weeks to months (13, 5). SSEs can be large, with a moment equivalent to moment magnitude (Mw) > 6 earthquakes (1, 2), and in some cases may precede and trigger damaging interplate earthquakes (7, 10). Slow earthquakes are observed in a wide range of geologic and tectonic settings but occur most commonly at subduction plate boundaries (1, 3). Despite their importance for fault dynamics and associated earthquake and tsunami hazards, the patterns of strain accumulation and release in slow earthquakes, their relationship to large, damaging earthquakes, and the underlying physical mechanisms that dictate observed variations in fault slip behavior all remain poorly understood. This is primarily due to the lack of direct constraints on in situ conditions and rock properties, as well as a dearth of continuous observations to resolve deformation offshore near the trench (2, 3, 7, 11).

Slow earthquakes are most commonly documented near the lower edge of the seismogenic zone that fails episodically in earthquakes (25, 9). Observations of SSEs and related phenomena updip of a large, locked seismogenic zone are far less common. Shallow SSEs have been detected at a number of subduction zones (2, 1217), some possibly reaching the trench itself, primarily at margins where the plate boundary appears to be mostly creeping (2, 12) or where the seismogenic zone is extremely heterogeneous (13, 14, 16). However, there is minimal constraint on shallow SSE recurrence or the distribution of slip, especially in the shallowest areas closest to the trench where onshore geodetic networks lack resolution. Although a few isolated VLFE swarms and LFTs have been observed in this environment and identified as thrusting events along the plate interface or within the upper plate (6, 7, 12), the extent of such events and their connection to patterns of fault locking and slip remain unknown. Among the outstanding questions are whether SSEs are commonplace on the shallowest reaches of subduction megathrusts and how they are related to VLFEs, LFTs, and large, damaging earthquake ruptures (7). Similarly, little is known about their role in accommodating long-term plate motion, particularly in strain accumulation and release near the trench where seafloor deformation is intimately linked to tsunamigenesis (18). Because the shallowest megathrust is accessible by drilling and high-resolution imaging, documentation of slow earthquakes there presents an important opportunity to characterize the in situ conditions and rock properties and thus to probe the underlying mechanics of these events (11).

We report on observations from an array of borehole and seafloor instruments at the Nankai Trough offshore of Honshu, Japan. Our instruments captured a series of repeating SSEs and associated tremor swarms over a ~6-year period from 2011 to 2016. The Nankai Trough is formed by northwestward subduction of the Philippine Sea plate beneath the Eurasian plate at ~6.5 cm/year (Fig. 1) (19, 20). Over most of its depth extent, the plate interface is interseismically locked, as defined by geodetic studies and a history of recurring large earthquakes and tsunamis, including the 1944 Mw 8.1 Tonankai event in which coseismic rupture extended to within 30 to 40 km of the trench (20, 21). Previous work in this region has documented multiple VLFE and LFT swarms between the trench and the shallow termination of estimated coseismic rupture in the Tonankai earthquake (6, 7, 12), raising the possibility that slow slip may occur updip of the seismogenic zone.

Fig. 1 Map and seismic cross section showing borehole observatories and seafloor sensor network.

(A) Map showing the DONET network (brown line; seafloor sensors denoted by brown triangles) and IODP boreholes (blue, C0002G; red, C0010A). Black contours (0.5-m interval) and the large black star indicate the rupture area and epicenter of the 1944 Tonankai earthquake (21). The epicenter of the Mw 6.0 earthquake on 1 April 2016 (35) is indicated by the smaller star. The dashed line shows the location of the seismic section in (B). (Inset) Zoomed-out view of area. (B) Seismic line showing NanTroSEIZE drill sites. White circles indicate the depth of pore-pressure monitoring intervals at the observatory boreholes C0002G (931 to 980 m) and C0010A (389 to 407 m). bsf, below seafloor; VE, vertical exaggeration.

The Integrated Ocean Drilling Program and the International Ocean Discovery Program (IODP) NanTroSEIZE (Nankai Trough Seismogenic Zone Experiment) project installed subseafloor borehole observatories at two drill sites above the plate interface (holes C0002G and C0010A) (Fig. 1 and fig. S1) (2224). The two holes are separated by 11 km in the dip direction and straddle the shallow limit of estimated coseismic slip in the Tonankai earthquake. We report pore-fluid pressure measurements within isolated depth intervals containing fine-grained and low-permeability sediments at the two observatories. Variations in formation pore pressure in these sealed intervals provide a robust proxy for volumetric strain (16, 17, 24). The borehole observatories are part of an extensive offshore observatory, the Dense Oceanfloor Network System for Earthquake and Tsunamis (DONET) (Fig. 1A), which includes a dense array of seafloor seismometers and pressure gauges linked to shore by a submarine cable (25). The observatory facilitates the detection and location of LFTs and VLFEs offshore (26) (fig. S2).

We found eight strain transients from the formation pore-pressure records (table S1). The strain transients included a variety of styles and event durations, all of which were caused by regional deformation, as they were synchronous at the two sites. The events lasted from a few days to several weeks, and some were accompanied by LFT swarms (Figs. 2 and 3 and fig. S3). The patterns of strain included: (i) mixed polarity, with a dilatation of ~0.05 to 0.24 microstrain (μstrain) (pore-pressure decrease of 0.26 to 1.35 kPa) at hole C0002G and a contraction of similar magnitude (0.13 to 0.15 μstrain, pressure increase of 0.62 to 0.71 kPa) at hole C0010A (four events); (ii) contraction at both sites (0.04 to 0.07 μstrain, corresponding to pressure increases of 0.22 to 0.34 kPa) (two events); and (iii) dilatation at both sites (0.38 to 0.5 μstrain; pressure decreases of 1.8 to 2.72 kPa) (two events) (Fig. 2B and fig. S3A). One event began with a mixed-polarity signal (October 2015) (Fig. 2C), but after ~11 days the sign of strain at the more trenchward borehole (hole C0010A) reversed. Both sites were then characterized by dilatation for the remaining ~3-week duration of the mixed event.

The character of the two exclusively dilatation events is distinct from the mixed-polarity and compressional events, in that the volumetric strains are larger (Fig. 3), and the events immediately follow the Mw 9.1 Tohoku-Oki (March 2011), Mw 6.0 Mie-ken Nanto-Oki, and Mw 7.0 Kumamoto (April 2016) earthquakes. This suggests that the SSEs were triggered by either static stress changes or dynamic stress changes from passing seismic waves. Although we cannot rule out the possibility of static triggering, dynamic triggering is most consistent with the large distances from the hypocenters of the Tohoku-Oki and Kumamoto earthquakes, the expected larger magnitude of dynamic relative to static stress changes, and widely observed magnitude-distance thresholds for earthquake triggering (27, 28). Triggering is well documented for normal earthquakes and tremors (1, 3, 27, 28), but observations of triggered SSEs are comparably rare (29).

Fig. 2 Formation pore-pressure and LFT records for three example SSEs.

(A) March 2014, (B) April 2016, and (C) October 2015. In all examples, the top panel shows pore-pressure records (red, hole C0010A; blue, hole C0002G) filtered to remove oceanographic signals (24). Thin trace lines in the top panels of (A) and (C) show smoothed data using a 12-hour window. The middle panels show the seismic moment release rate (Mo) [in Newton meters per day (Nm/day)] for LFTs, binned by color to indicate position relative to hole C0002G (distances as shown in maps below). Migration of LFTs is suggested in both (B) and (C) (primary moment release shifts to shallower depths, shown by warmer colors, over the duration of the swarms). Bottom panels show locations and magnitudes of LFTs (colors same as in middle panels, white for outside the square area), as well as DONET seafloor stations used to locate LFTs (gray triangles). For the April 2016 event (B), abrupt dilatation (decrease in pressure) initiates at the time of the Kumamoto earthquake; however, the preceding periods of rapid dilatation on 7 and 10 April do not correspond to any regional Mw > 4.0 earthquakes.

Fig. 3 Summary of pressure and strain transients at holes C0002G (blue) and C0010A (red).

Dashed vertical lines indicate the duration of each event. Those characterized by pressure increase (compressional strain; solid circles) at both boreholes are most consistent with 1 to 2 cm of slip centered >36 km from the trench, as shown in schematic cross sections for each event (compare with fig. S4A). Those characterized by dilatation (extension; open squares) at hole C0002G and compression at hole C0010A are most consistent with 1 to 2 cm of slip centered 24 to 38 km from the trench, and those by dilatation at both boreholes are consistent with slip of 2 to 4 cm centered <27 km from the trench. The October 2015 event (denoted by the asterisk) exhibited extension only at hole C0002G, and initial compression followed by dilatation at hole C0010A. εv, volumetric strain; ΔP, change in pressure.

Low-frequency tremors accompany the strain events but are restricted primarily to one region trenchward of both drill sites and a second region ~20 km along strike to the northeast (Fig. 2 and fig. S3). The locations of these LFTs are broadly consistent with those of well-located LFTs and VLFEs in the region from previous studies and clearly extend to the trench (Fig. 2 and fig. S3) (6, 7, 12). Some of the borehole pore-pressure transients are associated with energetic LFT swarms (e.g., October 2015 and April 2016), whereas others are accompanied by weaker activity. In most cases, LFT activity is generally concurrent with the strain transients in the boreholes (Fig. 2, A and B, and fig. S3). However, in some events (December 2012 and October 2015) (Fig. 2C and fig. S3C), LFT swarms initiate days after the strain transient begins. In the October 2015 case, the strain event observed in the boreholes begins on 10 October, whereas the LFT swarm initiates on 21 October—at the time of the reversal in sign of the strain signal at hole C0010A—and is localized mainly seaward of the borehole (Fig. 2C).

The concurrent LFT swarms and strain events at the two boreholes, together with their quasi-regular recurrence, suggest that they represent SSEs on the subduction plate interface below (6, 7, 16, 17). To explore this interpretation, we use a simple dislocation model to define the source locations, slip-patch widths, and fault displacements on the interface that are most consistent with the borehole data (24, 30). We can explain the observed patterns of strain with ~1 to 4 cm of slip on the megathrust occurring on a 20- to 40-km-wide patch, for all eight events. The polarity of the strain signal is dictated by the position of the patch relative to the boreholes (Fig. 3 and fig. S4). The best fit for mixed-polarity events requires 1 to 2 cm of slip centered below and between the two sites (25 to 36 km from the trench, ~4 to 6 km below seafloor). When contraction occurred at both sites, we found that a similar slip was required but needed to be centered farther landward (>36 km from the trench, >~6 to 7 km below seafloor). The dilatational signals potentially triggered by regional earthquakes required a slip patch centered seaward of both sites (<24 km from the trench, <~4 km depth) and larger slip (2 to 4 cm) relative to the other events (Fig. 3 and fig. S4). For the October 2015 event, the magnitude of strain and its reversal in sign partway through the sequence (Fig. 2C) are most consistent with a net slip of 2 to 4 cm and migration of slip from a region between the two boreholes toward the trench over ~3 weeks. The delayed onset of LFTs during the event and its localization to the region seaward of hole C0010A indicate that this SSE initiated silently and probably activated LFTs only upon reaching the area closer to the trench (Fig. 2C).

Based on our data, we suggest that periodic SSEs occur along the outermost reaches of the megathrust every 8 to 15 months (Fig. 3). Because the strain transients appear to repeat at relatively short intervals and the required deformation at the borehole sites is equivalent to that caused by centimeters of slip on the plate interface, these SSEs most plausibly reflect slip on the megathrust. If the deformation were distributed elsewhere, we cannot reconcile the required displacement with long-term slip on the plate boundary of ~6.5 cm/year. Additionally, the signs of strain at the two boreholes, in combination with the locations of LFTs (Fig. 2), are not well explained by slip on normal faults or on the megasplay fault in the upper plate, although these modes of deformation have been suggested for isolated events on the basis of inferred seafloor displacements and the focal mechanisms of VLFEs following the Tohoku-Oki earthquake (26, 31).

The fault slip amounts we determined from dislocation models require that these SSEs accommodate ~30 to 55% (~11 to 22 cm) of the total plate convergence budget (~39 cm) over the 6-year observation period. Slip rates for the SSEs, based on the estimated ranges of total displacement and the duration of each event, are ~0.1 to 2 cm/day, comparable to those documented for isolated shallow SSEs observed on other subduction megathrusts (2, 11, 13, 16). The two likely triggered SSEs exhibited larger slip and were centered farther seaward than the six spontaneous ones (Fig. 3 and fig. S4).

Our observations also allow us to illuminate the spatiotemporal relationships between SSEs and LFTs. The location and timing of primary slip we deduced from strain at the boreholes were not always coincident with LFTs. All of the spontaneous SSEs exhibited either compression at both boreholes or mixed polarity, indicating that the main slipping patch lies landward of hole C0010A. Yet LFT swarms were restricted mainly to the area 10 to 20 km trenchward of the boreholes (Fig. 2 and fig. S3). In comparison, events that exhibited primary slip nearest the trench (e.g., April 2016) were accompanied by many LFTs in that same region (Fig. 2, B and C). In the October 2015 event, the onset of LFTs at the time of strain sign reversal and the inferred migration of slip updip of hole C0010A provide additional evidence that LFTs do not fully map the slipping region. Instead, tremor appears to be activated by the SSE (rather than slip being driven by cascading tremor), and much of the slip occurs without any radiated seismic energy. The total seismic moment from tremor is on the order of 5 × 1013 Newton meters (N∙·m) (Fig. 2C), whereas the moment expected for 2 to 4 cm of slip in the tremor patch of area 400 to 1600 km2 (rectangular patch of 20- to 40-km dimension) would be ~4 × 1016 to 3.2 × 1017 N∙·m (Mw 5.0 to 5.6), assuming a rigidity of 10 GPa for the near-trench region. Finally, the recurrence of LFTs restricted to particular areas (near the trench and in a patch along strike to the northeast) (Fig. 2 and fig. S3) suggests that their occurrence is controlled by the presence of small-scale asperities, likely related to local fault frictional, compositional, hydraulic, or geometric properties that persist over multiple SSE cycles (32).

Our data provide evidence for repeated SSEs associated with tremor in the outer reaches of a subduction megathrust. The SSE source areas coincide with locations of previously observed LFTs and VLFEs (6, 7), within a region of the megathrust characterized by anomalously low seismic velocity and elevated pore-fluid pressure. This picture is consistent with models suggesting that SSEs and slow earthquakes originate in areas of very low effective stress (11, 33). The SSEs are arguably the clearest example of slow slip directly trenchward of a strongly locked seismogenic zone, along a megathrust that produces large earthquakes, raising the possibility that SSEs may also be common at similar subduction zones that host large or giant earthquakes, such as Cascadia (2, 15).

The regularity and relatively short recurrence of these SSEs, taken together with evidence for dynamic triggering, suggest that the outer subduction megathrust is mechanically weak and highly sensitive to perturbations. The SSEs account for a large proportion of the total plate convergence budget over the 6-year monitoring period. If this pattern is sustained over longer time scales, it implies that the shallow portion of the megathrust is unlikely to store substantial interseismic strain between large subduction earthquakes, consistent with relatively weak interplate coupling in this area over decadal time scales deduced from geodetic data (20). If these newly observed near-trench SSEs are a mechanism for periodic and frequent release of accumulated strain, the hazards associated with shallow earthquake rupture and tsunami generation in this region may be reduced. Continued offshore monitoring is critical for revealing longer-term patterns of strain accumulation and release, as well as possible links between shallow SSEs and large, damaging earthquakes (34).

Supplementary Materials

Materials and Methods

Figs. S1 to S4

Table S1

References (3638)

List of IODP Expedition 365 shipboard scientists

References and Notes

  1. Materials and methods are available as supplementary materials.
Acknowledgments: This work was supported by Japan Society for the Promotion of Science KAKENHI grant JP15H05717, NSF grants NSF-OCE 0623633 and NSF-OCE 1334436, and Deutsche Forschungsgemeinschaft grant KO2109/25-1. The borehole pore-fluid pressure data used for this research are openly available from JAMSTEC ( and The Pennsylvania State University (; the data set is also available from the National Environmental Information Center. Seafloor seismometer data are available from JAMSTEC on request.

Stay Connected to Science

Navigate This Article