Aseismic Transient Slip on the Gofar Transform Fault, East Pacific Rise

Home , Transform fault

Aseismic transient slip on the Gofar transform fault, East Paciﬁc Rise Yajing Liua,b,1, Jeffrey J. McGuireb,c , and Mark D. Behnb,d

aDepartment of Earth and Planetary Sciences, McGill University, Montreal,´ QC, Canada H3A 0E8; bDepartment of Geology and Geophysics, Woods Hole Oceanographic Institution, Woods Hole, MA 02543; cEarthquake Science Center, US Geological Survey, Moffett Field, CA 94035; and dDepartment of Earth and Environmental Sciences, Boston College, Chestnut Hill, MA 02467

Edited by Goran Ekstrom, Lamont-Doherty Earth Observatory of Columbia University, Palisades, NY, and approved March 19, 2020 (received for review August 6, 2019)

Oceanic transform faults display a unique combination of seismic that the material properties of the barrier zone are distinct from and aseismic slip behavior, including a large globally averaged those in the M ∼ 6 rupture zones. Prior to the 2008 Mw 6.0 main- seismic deficit, and the local occurrence of repeating magni- shock, the central barrier zone experienced ∼22,000 foreshocks tude (M) ∼ 6 earthquakes with abundant foreshocks and seismic (M0 to 4.5) extending in depth to the uppermost mantle, likely swarms, as on the Gofar transform of the East Pacific Rise and the a few kilometers deeper than the lower limit of oceanic crustal Blanco Ridge in the northeast Pacific Ocean. However, the under- seismicity defined by the 600 oC isotherm (10). An active-source lying mechanisms that govern the partitioning between seismic refraction line across the barrier zone revealed ∼10 to 20% and aseismic slip and their interaction remain unclear. Here we reduction in P-wave velocities in a several-kilometer-wide dam- present a numerical modeling study of earthquake sequences age zone extending through the entire oceanic crust (5), in con- and aseismic transient slip on oceanic transform faults. In the trast to a nearly intact fault zone in the fully coupled mainshock model, strong dilatancy strengthening, supported by seismic area (4). Systematically lower static stress drop values are also imaging that indicates enhanced fluid-filled porosity and possible estimated for earthquakes in the barrier zone (11), which are hydrothermal circulation down to the brittle–ductile transition, consistent with the seismic velocity reduction. Significant (up to effectively stabilizes along-strike seismic rupture propagation and ∼50%) seismic velocity reduction is also common along conti- results in rupture barriers where aseismic transients arise episod- nental transform faults and is usually attributed to a mechanical ically. The modeled slow slip migrates along the barrier zones damage zone due to intense fracturing (12). The depth extent of at speeds ∼10 to 600 m/h, spatiotemporally correlated with the continental low-velocity zones is less clear and it may not extend observed migration of seismic swarms on the Gofar transform. all of the way to the brittle–ductile transition as observed on the Our model thus suggests the possible prevalence of episodic aseis- Gofar fault (5). This large velocity anomaly on Gofar can be mic transients in M ∼ 6 rupture barrier zones that host active achieved by either enhanced fluid-filled porosity of 1.5 to 8% or swarms on oceanic transform faults and provides candidates a large volume fraction of mineral alternation at crustal depths for future seafloor geodesy experiments to verify the relation (e.g., 50 to 90% altered peridotite or >70% serpentinized troc- between aseismic fault slip, earthquake swarms, and fault zone tolite) (5). However, such large-scale alternation would produce hydromechanical properties.

oceanic transform faults | earthquake rupture segmentation | Significance aseismic transients | seismic swarms Oceanic transform faults slip through a unique combination of seismic and aseismic behavior. But the geological conditions veraged globally, oceanic transform faults (OTFs) release and physical mechanisms that lead to different slip modes on a small percentage (∼15%) of their accumulated moment A different parts of the same fault remain unresolved. Recent seismically, and despite the large thermally defined potential seafloor experiments document abundant microseismicity in rupture area the largest observed earthquakes have only mod- barrier zones that separate large earthquake ruptures. These erate magnitudes (up to magnitude 7 [M7]) (1, 2). Recent barriers require an explanation that permits small earthquake ocean bottom seismometer (OBS) deployment experiments nucleation, but prohibits large rupture propagation. We find along OTFs associated with fast-spreading ridges provide near- that coupling frictional dilatancy to a physics-based earth- field seismic data that shed light on the source mechanism quake cycle model can simultaneously reproduce the rupture of OTF earthquake rupture patterns and possible interaction segmentation and slip partition observed on the Gofar trans- between aseismic and seismic slip modes (3–7). On the Gofar form fault, East Pacific Rise. Our model shows a prevalence transform of the East Pacific Rise (EPR), M ∼ 6 earthquakes of slow slip events that drive migrating microearthquake repeat quasi-periodically every 5 to 6 y on 15- to 20-km along- swarms, a prediction that can be tested with future seafloor strike segments separated by a ∼10-km barrier zone (8). The geodesy experiments. rupture pattern is inferred from the 2008 foreshock and after-

shock distributions (3) (Fig. 1) and the relative centroid locations Author contributions: Y.L. designed research; Y.L. and J.J.M. performed research; Y.L. and of M6.0+ events on Gofar since 1992 (2, 8). Similar earthquake J.J.M. analyzed data; and Y.L., J.J.M., and M.D.B. wrote the paper.y rupture patterns are observed along the Blanco Ridge trans- The authors declare no competing interest.y form of the northeast Paciﬁc Ocean, where a ∼10-km barrier This article is a PNAS Direct Submission.y zone separates two distinctive moment magnitude (M ) 6.0 to w Published under the PNAS license.y 6.5 patches that rupture quasi-periodically every ∼14 y (2, 7). Data deposition: Seismic waveform data from the 2008 Gofar ocean bottom seismome- Repeated ruptures are also reported on transforms that offset ter experiment are archived at the Incorporated Research Institutions for Seismology slow-spreading ridges, such as the Charlie-Gibbs transform of (IRIS) data center, network code ZD, https://www.fdsn.org/networks/detail/ZD 2007/. the Mid-Atlantic Ridge (9), although large location uncertainties Data products can be accessed on Open Science Framework at https://osf.io/rw9bq/.y exist using teleseismic data. 1 To whom correspondence may be addressed. Email: [email protected] The persistent rupture barrier through multiple cycles, recent This article contains supporting information online at https://www.pnas.org/lookup/suppl/ OBS-based observations of the ﬁne-scale seismicity distribution doi:10.1073/pnas.1913625117/-/DCSupplemental.y (3), and fault zone velocity structure on Gofar (5) jointly suggest First published April 28, 2020.

10188–10194 | PNAS | May 12, 2020 | vol. 117 | no. 19 www.pnas.org/cgi/doi/10.1073/pnas.1913625117 Downloaded by guest on October 1, 2021 Downloaded by guest on October 1, 2021 i tal. et Liu transi- stability (friction lines white two tion, within Area Gofar. western senting (a ( parameter (3). al. stability et McGuire friction from modified credit: Image M locations. 2008 troid the of M extents 2007 tri- rupture and estimated White (west) the 8. show to ellipses 7 White strong-motion December sensor. with colocated on seismometer seismicity stars, on white swarm aftershocks seismometer; red, cyan, angles, 12; 20; to to 10 18 September September on foreshocks Yellow, 2008. M August 18, an in ber surrounding located dots), and (black recorded 2008 was December Seismicity to (A) setup. model cycle quake 1. Fig. a (a strengthening ity . utr n h are oe(i.1,wihsailycoin- M spatially offset which trace 1), fault (Fig. a zone with barrier 2008 cides the the and between boundary rupture the 6.0 at offset trace fault ∼600-m zone barrier the in Gofar. reduction on velocity seismic the primary sensitivity the for likely measurement mechanism is porosity the fluid-filled below enhanced Therefore, the signature (5). across gravity observed would a change that porosity produce estimated than the whereas greater zone, barrier much Gofar anomaly gravity a smr ai hnabetfliscnflwit h e pore effective new and the reduced into temporally flow is can pressure fluids pore ambient local the grains, than space, consequently fracturing and rapid and rotating more rate, by space is sliding void shear new creating the of rate If 18). (17, behavior. high begins slip and fault porosity influence can fluid-filled which large of in both supply pressure, result pore persistent can The fluids 16). excess (7, slip of transform Ridge aseismic observed Blanco recently favor the where as along swarms, mantle, to earthquake abundant upper suggested promote the and is into serpentinization transition pervasive depth brittle–ductile the increasing the hence and of low- temperature for zone basin responsible fault pull-apart the likely a ering also of is that circulation fault to Hydrothermal similar the (15). and manner in a faulting in normal complexity infiltration, for local fluid conditions such favorable 14), provides (13, geometry faults transform tal C B A −4.4˚ −4.6˚ − 164 162 16 158 −105.6˚ −105.8˚ −106˚ −106.2˚ −106.4˚ w rcs esiiyrlcto nwsenGfrietfida identified Gofar western on relocation seismicity Precise nfli-auae rnlrmtras ooiyeovsa shear as evolves porosity materials, granular fluid-saturated In Depth (km) 10 b, 4 3 2 1 02 040 30 20 10 0 −10 −20 −30 −40 5 0 a . yai utr ae nosrain rmcontinen- from observations on based rupture dynamic 6.0 Depth (km) 10 hrceitcsi distance slip characteristic a, − 5 0 oa rnfr al esiiy 6rpuesget,adearth- and segments, rupture M6 seismicity, fault transform Gofar iaac dilatancy dilatancy b 0 = a−b )hsvlct-eknn rcin(a friction velocity-weakening has 0) 0.02 0--epsaordpeso 4.Atog a Although (4). depression seafloor ∼500-m-deep 0.04 w − . es) ihga imnsdntn epciecen- respective denoting diamonds gray with (east), 6.2 10 mi o ieyt esfcett tpa stop to sufficient be to likely not is km <1 5 0

0 M6 rupture M6 > 010 ) (C 0). 0.02 a Along−strike (km) − 0.04 dilatancy et itiuin ffito parameters friction of distributions Depth ) apdo h rnfr al repre- fault transform the on mapped b) KM d

10 5 0 “Barrier” c n fetv omlstress normal effective and , 0 20 d c (mm) 20

30 M6 rupture M6 40 w 10 − 5 0 anhc nSeptem- on mainshock 6 0 b < 20 (MPa) ) usd sveloc- is outside 0); 40 60 Rate–state B) σ ¯ . (a b) 0 M 0 0.02 0.04 w w 6 M The above. discussed of occurrences as unlikely appears serpentinization vasive eki hr 0k al emn 3.Teaudn and strongly abundant zone mainshock The M6 the (3). to Gofar adjacency western its segment and the zone fault in barrier microseismicity 10-km of distribution short widespread a in spectacularly per is week microearthquakes zone of thousands barrier generating Gofar active, at micro- western seismically area the of fault contrast, the streaks friction of By the majority VS. VW along the areas leaving section, therefore small (28), creeping to seismicity (27). limited at Fault be Fault ruptures to Andreas Andreas appears San earthquake San central the Parkfield the Within along of section best arrest creeping as the segments, the adjacent by from m ruptures illustrated VW–VS M6 largely a stop a to Such in effectively embedded background. patches creeping VW VS, on earthquakes small erate of distributions stochastic conditions. VS under aseismically than zone lower fault largely because a unlikely is is there zone VS temperature crustal-scale a the such in ever, other VS as all central S2; same Fig. the the Appendix, into (SI parameters barrier propagate rupture not M6 a in do zones, resulting VW zones zone, 20-km VW two the along- between in 10-km embedded ruptures a is when that zone find VS we strike Gofar, neighboring For the (26). from segments originated VW that ruptures seismic of agation et,alre Szn,hge fetv omlsrs,adlarger and stress, normal a effective seg- higher zone, VS/VW numeri- VS alternating larger in a homogeneous, ments, example, of For consisting zones. models VW cal the are ruptures within earthquake confined of that mainly such distributions parameters, spatial model design other to with used be can erties IAppendix, earth- single (SI a mechanism in S1 stabilizing ruptures Fig. effective zone VW an depth) without in entire quake km the 5 that to the such (1 here, satisfied ∼4-km always presented is model the condition the on instability In conditions above ). Methods pressure and pore the (Materials and by of fault stress, defined dimension normal size down-dip nucleation friction, characteristic the the a when exceeds achieved segment be VW Seis- can (10). rheology slip visco-plastic mic and calculated structure cooling thermal hydrothermal Gofar with western a using (25) data friction rate–state gouge model, gabbro temperature-dependent our converting In Methods). seismic and a into (a grow to ble potential the (a has rate–state event slip the whether order In parameter first (22). stability friction laws the friction on framework, state-dependent sequences laboratory-derived and by rupture governed rate- segment earthquake fault Gofar simulate western the to bound- conditions, and loading ary OTF-specific including on transform, based Gofar model the numerical (3D) three-dimensional a Dilatancy develop We with Coupled Model Sequence Earthquake OTF transform oceanic the for budget. implications slip fault its the on and behavior transform barrier rupture Gofar persistent the for strengthening in mechanism dilatancy a observational of as recently aforementioned feasibility the the more investigate on we Based evidence, and 21). subduc- (20, (17–19) in zones landslides, events tion slip gouges slow in episodic fault generating deformation models numerical and dynamic tills, inhibiting to glacier in proposed effective been enhanced. has is be pressure) mechanism pore dilatancy-strengthening minus This stress (normal stress normal − − w oecmlxsailhtrgniis uha admor random as such heterogeneities, spatial complex More h tbeusal optto ittdb h SV prop- VS–VW the by dictated competition stable–unstable The . losgetta h nieoenccutcno ipyslip simply cannot crust oceanic entire the that suggest also 6.0 b b > ausaealfvrbecniin o niiigteprop- the inhibiting for conditions favorable all are values − ). 0 b − ae nlbrtr abofito aa(5 n per- and (25) data friction gabbro laboratory on based > b eoiysrnteig[S)(3 4 (Materials 24) (23, [VS]) strengthening velocity 0, < 000frsok ntebrirzn ote2008 the to zone barrier the in foreshocks ∼20,000 eoiywaeig[W)o ilrmi sta- remain will or [VW]) weakening velocity 0, PNAS | a a 2 2020 12, May − a b i.S1 Fig. Appendix, SI a lob nrdcdt gen- to introduced be also can , − b smpe nteOFby OTF the on mapped is | o.117 vol. ∼500 a − ´ lnecnwork can elange b ◦ | eurdfor required C ittst the to dictates o 19 no. a ae.How- case). − b along , | 10189

EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES suggest, if such a VW–VS mélange is present, that the VW sure evolution, controlled by frictional slip along the fault and patches predominate and are separated only by sporadic VS fluid diffusion across the fault, is coupled to the rate–state fric- patches. Despite this predominance of VW material, the bar- tion framework to obtain the spatiotemporal evolution of slip rier zone repeatedly stops large ruptures (3). This proportionality rate, shear stress, and pore pressure on the entire fault (19, 20) would be analogous to the stochastic VW–VS distribution in the (Materials and Methods). transition zone at the bottom of the seismogenic layer proposed to reconcile seismicity and geodetic locking depths on the Anza Rupture Barrier section of the San Jacinto Fault (29). As numerically demon- Fig. 2 A and B shows a ∼10-y history of slip and slip rate for strated for Anza (29), under such a frictional stability parameter a simulation case with porosity φ = 0.05 and dilatancy coeffi- distribution, an earthquake rupture nucleated from the large cient = 10−4, which are in the range of parameters estimated VW seismogenic zone would still propagate through the VW–VS for western Gofar (see SI Appendix, Figs. S1 and S3, for addi- mixture, inconsistent with the rupture barrier observation here tional simulation cases with other values of φ and ). In contrast on Gofar. to a whole-fault rupture as predicted with no or a weak dila- Thus to construct a model consistent with these field obser- tancy effect ( = 10−5; SI Appendix, Fig. S1), earthquake rupture vations and numerical predictions in previous studies, dilatancy in the model with zones of strong dilatancy is separated into strengthening is introduced on three along-strike segments of western and eastern segments, repeating every 5 to 6 y (Fig. 2 the fault, 10 km at the center and 15 km on each end; fault A and B). Coseismic rupture initiates at the boundaries of properties elsewhere are not modified (Fig. 1B). Such an along- strong-dilatancy to no-dilatancy zones, and propagation is mainly strike distribution of dilatancy is chosen to spatially reflect the contained within the no-dilatancy zone with quickly diminishing M ∼ 6.0 rupture patterns on western Gofar (Fig. 1A), which rupture speeds approaching the boundaries on either side. As a allows us to determine whether dilatancy can effectively stop the result, the central dilatancy zone (−5 to 5 km) becomes a rupture ruptures given the plausible porosity values inferred from the barrier to M ∼ 6 events on each side. The other two dilatancy P-wave anomaly in the barrier zone (5). There are two key dila- zones at the western and eastern ends of the fault also remain tancy parameters in our model setting. The first parameter is the aseismic. The pair of modeled M ∼ 6 ruptures follows a gen- nondimensional dilatancy coefficient , which measures porosity eral pattern where the seismic cycles of the two rupture patches changes with shear velocity steps. A higher corresponds to a have similar phasing such that the M ∼ 6 ruptures often hap- larger increase in porosity for a given increase in sliding veloc- pen within a few months of each other, similar to the westward ity and hence greater tendency for slip stabilization and arrest; sequence of rupture observed on Gofar in the past earthquake is between 10−5 and 10−4 measured in fault sliding experiments cycles (8). on natural and simulated gouges (19, 30, 31). The second param- We further quantify the seismic coupling ratio (χs ) within the eter is the bulk porosity φ, which varies between 0.01 and 0.1 central barrier zone as a function of the degree of dilatancy, as inferred for the Gofar rupture barrier zone (5). Pore pres- represented by the nondimensional parameter /φ (Fig. 2C; see

AB3 20

18 2.5 16

14 2 EQ2 EQ1 EQ2 Slip (m) Time (yr) 12 EQ1 SSE2 1.5 SSE2 SSE1 SSE1 10

-40 -20 0 20 40 10 5 0 Along-strike (km) Log (V /V ) CD10 max pl 0.3 1 =0.01 =0.02 0.25 =0.05 2a, 2b

0.8 =0.1 t s 0.2 0.6 0.15

0.4 0.1 Seismic Transient

0.2 2a, 2b 0.05

0 0 10-4 10-3 10-2 10-1 10-4 10-3 10-2 10-1 / /

Fig. 2. Modeled earthquake rupture pattern and aseismic vs. seismic slip budget. (A) Cumulative slip in multiple earthquake cycles at depth 4 km. Black 6 6 solid lines are interseismic slip (Vmax < 10 Vpl) plotted every 0.5 y. Red dashed lines are coseismic slip (Vmax ≥ 10 Vpl) plotted every second. Plate loading rate Vpl = 14 cm/y. (B) Normalized maximum slip rate log10(Vmax /Vpl) on the fault for the same period as in A. SSE1 and SSE2 denote the timing and locations of two aseismic slip transients; EQ1 and EQ2 denote two earthquakes. φ = 0.05, = 10−4 for the case in A and B.(C) Fraction of slip released during seismic events χs, averaged over the modeled earthquake cycles and within the central high dilatancy zone (−5 to 5 km) under different dilatancy conditions, represented by the ratio between dilatancy coefﬁcient and porosity /φ.(D) Fraction of aseismic transient slip χt averaged over multiple earthquake cycles and within the central high-dilatancy zone. Black arrows in C and D point to the simulation case shown in A and B.

10190 | www.pnas.org/cgi/doi/10.1073/pnas.1913625117 Liu et al. Downloaded by guest on October 1, 2021 Downloaded by guest on October 1, 2021 ic rmteso lpeet bevdi udcinzones subduction in observed and events Dis- faults slip barriers. slow transform rupture be the OTF oceanic may within from along they mechanism tinct slip slip suggest similar of prevalent swarms and a mode Gofar instrumen- model common the of numerical a to our lack pattern been in to migration have appearance due events their OTFs slip tation, along the transient geodetically for within no migration detected inferred Although wave as barrier. porosity of rupture such and manifestation (33), Gofar, a zones mech- subduction possibly on driving and (32) the swarms faults strike-slip be continental seismicity may for slip transient anism peak aseismic the that outside propose cases for quantify migra- dilatancy to clear χ difficult of fol- slip; are range d transient patterns the significant tion 2 average for produce The to much that period. vary 1 parameters remaining not first the does the speed for migration in speeds slower migration by faster lowed with curve, fusional (χ budget slip 2D. transient Fig. aseismic in peak the to sponds φ h rpgto fmdldaesi lptaset iharange a with coefficients transients dilatancy slip of aseismic modeled of by described propagation accurately the longitude be can 1A, segment envelope Fig. migration westernmost seismicity in The during the dots on episode (cyan experiment, Gofar swarm of OBS abundant Gofar most 2008 the the events, swarm sim- 2008 are cases Both with ). S4 Fig. ulated Appendix, (SI and zone within barrier 3) west-end m/h (Fig. 600 zone to barrier 10 central average of speeds speed Propagation d. a 10 at zones barrier rupture M locked of fraction budget. significant slip a OTF 25%, total to 15 the reach can zone barrier central 2A Fig. For in Methods). case and the (V (Materials for slip rate cycles total example, the earthquake loading to modeled full seismically tiple the released slip at of accumulated ratio the as defined Methods and Materials i tal. et Liu than lower or stronger S3 near a rates whereas slip transients to leads aseismic effect value to slip intermediate seismic stabilize an to in result that /φ pairs coefficient dilatancy 2C), (Fig. effect ratio coupling seismic the to χ contrast In Methods). phases) and postseismic (Materials and coseismic and nucleation earthquake (several ing threshold ratio detection slip geodetic the times transient exceed aseismic would above M that the magnitude the slip quantify of between we orders periods Here three 2 intervening to Fig. the one in ing rate illustrated the slip below, as in or events, at, slip simply aseismic slip not rate does loading plate zone barrier central The Swarms Seismic Driving Transients Slip Aseismic result and M slip to Gofar. seismic western barriers serve arresting rupture can for mechanism strengthening in effective dilatancy com- an that a as hypothesis to the prior increasing confirm with it decreases into ratio coupling as propagating /φ the zone, slip Overall, barrier stop. coseismic the plete in of seismically result released a is moment total the t s 0 = h simcsi rpgto rnsi eea olwadif- a follow general in fronts propagation slip aseismic The h rnin lpeet ulaea h eihrl fthe of peripherals the at nucleate events slip transient The .Wti h nemdaerneof range intermediate the Within ). hc oooial erae iha nraigdilatancy increasing an with decreases monotonically which , ag u olwrsi ae n mle lpaes We areas. slip smaller and rates slip lower to due range 0 = r qiaety rae iaac fet u eut thus results Our effect. dilatancy greater equivalently, or, oeta hsrneof range this that Note .05 V .002 pl u sntascae iherhuks(.. exclud- (i.e., earthquakes with associated not is but ) ∼ ;awae iaac feti o sufficient not is effect dilatancy weaker a 2D); (Fig. utr oe n rpgt costeadjacent the across propagate and zones rupture 6 10 = χ −4 V t pl sgets o h hie fprst and porosity of choices the for greatest is and ahr hsrgo xeine transient experiences region this Rather, . o h omlto of formulation the for (0.7 = φ ∼ 0 = atqae sosre along observed as earthquakes 6 i.4sosteDecember the shows 4 Fig. .05 /φ to 0mhfrteeet nthe in events the for m/h ∼20 0 / o h vnsi the in events the for m/h ∼100 (/φ 1.2) (1.4 = h average the /φ, × 0 = 10 pl 0. to −106.2 V . to ,aot3%of 30% about 002), 14 = −4 pl ihexcursions with B, 2.4) ∼ Fig. Appendix, (SI .Here /φ). n ooiyof porosity a and earthquakes. 6 my nmul- in cm/y) × t 10 −106.3 χ sshown as ) t −3 χ V sfault as t pl corre- to ∼5 nthe in χ dur- ◦ s S). is i.3. Fig. nteGfrDsoeyadBac ig rnfr faults transform Ridge Blanco and of Gofar-Discovery mixture the occurrences a quasi-periodic on The to (22). leads zone between slip of framework) nucleation seismic ratio rate–state and characteristic aseismic dimensional the the in and the hydrome- (defined fault zone or frictional in seismogenic changes (36) either the to as conditions due modes time transition chanical slip over 8). segments evolve seismic (3, fault properties and hydration scenario, of aseismic rupture degree between due multimode and properties the lithology occur frictional In variable distinct persistently of spatially earthquakes segments to essentially coupled scenario. scenario, faults fully rupture single-mode multimode on transform a the oceanic or single-mode In on a slip either involve seismic aseis- between and partitioning the mic explain to geodesy proposed Hypotheses seafloor Conclusions future and Discussion by hypothesis dilatancy the slip experiments. aseismic of each test in to up slip ments reach cumulative can maximum episode the 35), (34, along-strike in within shown defined coefficient period is Dilatancy 10.5-d km. zone 5 active high-dilatancy the central during The speeds A–D. migration front slip sient (t when y episode 0.42 in mulated between M rates in at ruptured slip transient migrating time ∼V represent starting reference colors a Red/orange from 2B. d 10.5 and 0.84, ( B).

pl Migration speed (m/hr) 1000 M

npht fnraie lprt ntefuta ie ,0.43, 0, times at fault the on rate slip normalized of Snapshots A–D) Depth (km) 10 10 10 10 10 200 400 600 800 n 10 and 5 0 5 0 5 0 5 0 0 5 w 4 2 2 0 -20 -40 0 oee simctasetsi rpgto SE nFg 2 Fig. in (SSE1 propagation slip transient aseismic Modeled D C E B A 5051 15 10 5 0 -5 F o6erhukswt iia etodlocations centroid similar with earthquakes 6 to 5 Day 10.5 4 V pl ,rneigaunique a rendering S6), Fig. Appendix, (SI cm ∼10 V ∼ akbu ra ersn okdsget htpreviously that segments locked represent areas blue Dark . max atqae.(E earthquakes. 6 > 02 o1.8y,dﬁe stedrto fthis of duration the as deﬁned y), 10.68 to 10.26 = 10 PNAS ,wt ufc displace- surface with 3E), (Fig. cm ∼30 0.8 V Along-strike (km) = pl Alon | (F . 10 a 2 2020 12, May slow slipmigration −4 etadadeswr simctran- aseismic eastward and Westward ) g 10.5 day 0.84 day 0.43 day -strike (km) Day 0 porosity , oa fmxmm3 msi accu- slip cm 30 maximum of total A ) 0.42 yr 0 day φ | = o.117 vol. t = 0.05. 04 ssoni Fig. in shown as y 10.5 Cumulative slip(m) | o 19 no. Log 10 0 (V/V | A to −5 pl 10191 ) -2 0 2 4 -2 0 2 4 -2 0 2 4 -2 0 2 4 and 0 0. 2

EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES -106.1 gates on average at ∼20 to 100 m/h over a period of about 1 wk, at least an order of magnitude slower than those on the Blanco transform fault. Several other swarm episodes are also identiﬁed with migration speeds <1,000 m/h from the relocated -106.15 epicenters (SI Appendix, Fig. S7) (4), although the event num- bers of each episode are far fewer than those in the December 2008 episode. Furthermore, the Blanco swarms demonstrate a -106.2 roughly linear migration pattern (7), whereas the majority of 100 m/hr the Gofar December 2008 swarm sequence followed an approx- imately diffusive curve (Fig. 4, solid lines). The Gofar swarm

Longitude -106.25 migration speed and pattern are indeed more comparable to those of seismic bursts in southern California exhibiting migration indicative of fluid diffusion (41). The above differences in the Gofar and Blanco swarm migration characteristics may -106.3 suggest different degrees of fluid involvement. Hydrothermal circulation and the minor extent of mineral alternation (10) com- bine to enhance porosity and hence dilatancy effect in the Gofar 338 340 342 344 346 348 rupture barrier zones, thereby promoting aseismic transient slip Day of Year 2008 and driving diffusional swarm migration at crustal depths. By contrast, pore space in the Blanco mantle layer may be primar- Fig. 4. Seismic swarm in December 2008 and modeled aseismic slip migra- ily filled by serpentine, such that little to no dilatancy effect tion fronts. Red dots show relocated M ∼1 to 3 earthquakes detected within ∼10 d near the Gofar transform–ridge interaction toward the west end. is exhibited nor are diffusional patterns of swarm migration Gray dots show matched-filter–based detections of smaller events (M0 to observed. 2) that were not relocated independently from their template (3). Black The presence of abundant aseismic slip, evolving in time dashed line shows a linear migration speed at 100 m/h, for reference. and space, has fundamental implications for the type of earth- Solid lines show modeled diffusive aseismic slip fronts from simulation cases quake sequences and total slip budget on major plate boundary using porosity φ = 0.05 and dilatancy coefficient = 0.7 × 10−4 (cyan, three faults. Future work is required to understand how aseismic slip −4 −4 −4 cases), 0.8 × 10 (blue, two cases), 10 (black, two cases), and 1.2 × 10 leads to seismic swarms, sometimes in the form of repeating (green, one case). microearthquakes, as has been reported in several cases pre- ceding large megathrust earthquakes (42, 43). The potential stress triggering between aseismic creep, seismic swarms, and (7, 8) are more consistent with the single-mode pattern. The subsequently large earthquakes may allow us to better under- OTF slip evolution modeled in our study after Gofar clearly stand the earthquake nucleation process that has long evaded demonstrates that strong dilatancy can stop seismic rupture seismologists. propagation and result in persistent rupture barrier zones as observed for the Gofar transform. For the Blanco Ridge trans- Materials and Methods form, fault geometry and bathymetry changes may play more Rate–Station Friction Coupled with Dilatancy. We implement the laboratory- important roles in the Mw 6.0 to 6.5 central barrier zone (7), derived rate- and state-dependent friction law (23), coupled with dilatancy where the occurrence of a local magnitude 5.5 indicates possi- (19), to simulate fault slip (δ), shear stress (τ), and pore pressure (p) bly higher fault strength and weaker dilatancy effect than those evolution in OTF earthquake cycles, following in the barrier zones along Gofar. Future studies on oceanic transform fault geometry, velocity structure, and high-precision τ =σ ¯f = (σ − p) [f0 + a ln(V/V0) + b ln(V0θ/dc)], [1] seismicity location are needed to determine to what extent the dθ Vθ dilatancy mechanism affects global OTFs’ earthquake rupture = 1 − , [2] patterns. dt dc dp p − p0 1 dθ Our modeling results also suggest a two-way stress trigger- = − + , [3] ing between aseismic creep in the high-dilatancy, rupture barrier dt tp β θ dt zones and the M ∼ 6 in the coupled seismogenic zones. Strain energy release from an aseismic transient slip event in the bar- where θ is the state variable. Effective normal stress σ¯ ≡ σ − p, and rate– rier zone increases stresses around the peripherals of the slip state friction parameters a, b, dc are time constant but spatially variable. −6 zone and promotes earthquake nucleation in the neighboring Reference slip rate V0 = 10 m/s, and nominal friction coefficient f0 = 0.6 segments. Coseismic and postseismic slip in the M ∼ 6 rup- when V = V0. The earthquake nucleates on a velocity-weakening (a − b < 0) fault of a dimension greater than the characteristic nucleation size defined ture zones consequently increases loading on the adjacent barrier as h* = γ µbdc (44), where µ is shear modulus, and γ is a geometrical fac- zones and leads to aseismic transients (Fig. 2 A and B). We note (b−a)2σ¯ tor of unity. We adopt laboratory measurement of temperature-dependent from Fig. 2B that not all aseismic transient slip events precede friction stability parameter a − b for gabbro gouge under hydrothermal an earthquake rupture, which suggests that the stress state of the conditions (25) and convert it to be depth-dependent (Fig. 1) using a future seismogenic zone must be taken into consideration. Gofar thermal model (10) that accounts for hydrothermal circulation effect. Seismic swarm activity has been long recognized on plate σ¯ = 50 MPa and the characteristic slip distance dc is ∼5 mm such that the boundary faults (37–39) and is suggested to be a manifestation width of the seismogenic zone (W ∼ 4 km) is about 3.5 times that of the of brittle failure on small asperities driven by aseismic creep nucleation zone h* ∼ 1.14 km, under which nearly all slip is released in in the crust or upper mantle layers (7, 40). Swarm migration earthquakes (22). characteristics thus provide useful insights into the underlying The elastic relation between slip δ and shear stress τ follows the “quasi- fault slip behavior. A recent study using ocean bottom seis- dynamic” approximation (45) mometers deployed near the Blanco Ridge found consistent Z X Z W swarm migration rates from 100 to 2,000 m/h with a median L d 0 0 0 0 0 0 τ(x, ξ, t)= − k(x − x , ξ, ξ ) δ(x , ξ , t) − Vplt dx dξ of 370 m/h (7). A wide range of swarm migration speed is 0 0 also found for western Gofar. For the December 2008 swarm ∂δ(x, ξ, t) − η , [4] episode on the west end of Gofar (Fig. 4), seismicity propa- ∂t

10192 | www.pnas.org/cgi/doi/10.1073/pnas.1913625117 Liu et al. Downloaded by guest on October 1, 2021 Downloaded by guest on October 1, 2021 4 .P is,S .Wsosy tp n asi rudrpue:Eprclbud on bounds Empirical ruptures: ground in gaps and Steps Wesnousky, G. S. Biasi, P. G. 14. temporal and Spatial Collins, A. J. McGuire, J. J. Boettcher, S. M. transform Moyer, oceanic A. P. of 11. behavior Thermal-mechanical Hirth, G. Behn, M. Roland, E. 10. 3 .G enuk,Simlgcladsrcua vlto fsrk-lpfaults. strike-slip of evolution structural and Seismological Wesnousky, G. S. 13. Cochran S. E. 12. .. xldn lpi atqaenceto n otesi hss is phases, postseismic and nucleation earthquake in slip earthquakes, with associated excluding not periods during i.e., slip aseismic of subset a and oqatf esi essaesi lp lpi enda esi ( seismic as defined is Slip slip. aseismic V versus seismic quantify to simulation, between each of Specif- mode. θ beginning slip nonuniform the into and at state ically, steady off evolve to system the evo- pressure pore ratio to parameters (Eq. controlling lution the on assumptions, focus assuming above we to the study equivalent refer- is this their this in effect; to Therefore, of similar condition (22). undrained are nearly dilatancy the earthquakes without modeled values of ence range intermediate depth effect. at rupture dilatancy that no found conditions, have to drainage studies equivalent mean- to numerical timescale, drained), return previous instantaneously rupture (effectively Furthermore, nearly coseismic (5), level would Gofar the ambient change of to pressure its pore zone comparable any barrier is that rupture which ing the s, for 100 inferred to as example, 0.1, for to conditions, 0.01 drainage intermediate At Pa·s. i tal. et Liu Budget. Slip OTF earthquakes. of cycle first θ compressibility fluid compressibility solid with defined compressibility, matrix stiffness The respectively. OTF, modeled x the of dimensions dmg oe 4)rslsi ierneof range wide a in results where (46) [s], zone) (damage m ∼1 of range plausible A thickness zone fault thickness finite of zone fault pressure pore zone time. (x (x strike location at slip where .K drod .E brrmi,Te21 w71erhuk nteCharlie-Gibbs the on earthquake 7.1 Mw 2015 The Abercrombie, E. R. Aderhold, transform K. Rise Pacific 9. East on predictability earthquake and at cycles interaction Seismic crust-mantle McGuire, and J. slip of 8. Mode Braunmiller, J. Nabelek, L. J. Kuna, M. V. between 7. relationship The Collins, J. McGuire, on J. constraints Boettcher, M. velocity Wofson-Schwehr, M. Seismic 6. Collins, J. McGuire, J. Lizarralde, D. Roland, E. 5. Froment B. 4. ridge mid-ocean on ridge McGuire cycles seismic mid-ocean J. for 3. relations for Scaling McGuire, relations J. J. scaling Boettcher, S. Earthquake M. Jordan, 2. H. T. Boettcher, S. M. 1. 0 and , max = , ml etrain r nrdcdit h nta odtost allow to conditions initial the into introduced are perturbations Small nEq. In utr propagation. rupture (1988). 340–343 335, Rise: Pacific East Fault, Transform Gofar on strength. drop fault for stress Implications earthquake in variations seismicity. of distribution spatial the oceanic Geosys. for slow Implications a faults: on slip multimodal and earthquakes transform. Repeating fault: transform faults. Rise. Pacific East faults. transform fault, oceanic transform Discovery the on Geosyst. Geophys. Quebrada- structure Geochem. fault the and at seismicity behavior Rise. earthquake Pacific control East (2012). faults, that transform properties Discovery-Gofar material Rise. the Pacific East fault, transform (2014). Gofar the Rise. Pacific East fault, transform faults. transform faults. transform aaezones. damage ξ d , ≥ ξ c /φ. /V 0 p ersnssersrs hnea oain(x location at change stress shear represents ) mm/s 5 X oacutfrtasainlsmer n euecomputation reduce and symmetry translational for account to ) ul eso.Sc Am. Soc. Seismol. Bull. pl L vltosbcm needn fteiiilcniin fe the after conditions initial the of independent become evolutions 5to −25 3, 001(2010). Q07001 11, = φ sed tt)eeyhr ntefut xetfrtesegment the for except fault, the on everywhere state) (steady are 3) epy.Rs Lett. Res. Geophys. t 0k and km 80 sfutzn ooiyand porosity zone fault is p aitosi atqaerpuepoete ln h Gofar the along properties rupture earthquake in Variations al., et mgn ln-tievrain nmcaia rprisof properties mechanical in variations along-strike Imaging al., et ≡ esi n edtceiec o xesv,ln-ie fault long-lived extensive, for evidence geodetic and Seismic al., et ≈ Geology ηβ euetemxmmfutsi rate slip fault maximum the use We maogstrike; along km −5 epy.Rs Lett. Res. Geophys. Res. Geophys. J. 10 0 , p d β 6 and ξ d 2 V t oreulbaewt h min pressure ambient the with reequilibrate to 0 ul eso.Sc Am. Soc. Seismol. Bull. φ k /k p ,weeaFuirtasomi ple ln the along applied is transform Fourier a where ), pl n permeability and /(d ≈ = 1–1 (2009). 315–318 37, a.Geosci. Nat. hc ne h eodtr nEq. in term second the enter which φ, W n simc(χ aseismic and stecaatrsi ifso iecl o fault for timescale diffusion characteristic the is 10 10 c d /V 6831 (2014). 3698–3712 15, −14 −2 0718 (2008). 1067–1084 98, = 1962 (2016). 6119–6128 43, d .Gohs Res. Geophys. J. pl 0k r h ln-tieadalong-dip and along-strike the are km 10 a.Geosci. Nat. 1) ti hlegn oestimate to challenging is It (19). MPa 132(2004). B12302 109, ) o10 to ≈ t 231(2009). L21301 36, p 1t .,tettlcsimcsi and slip coseismic total the 0.1, to 0.01 3–4 (2019). 138–142 12, −1 ≈ k −20 n uddnmcviscosity dynamic fluid and , 10 θ = β = a 1012 (2016). 1110–1124 106, 9 k 10 3–4 (2012). 336–341 5, = hn10 when ) m hc aiie h dilatancy the maximizes which φ, d r agl nnw o Gofar. for unknown largely are 7274 (2018). 7722–7740 123, 2 −20 c φ(β /(2.5V .Gohs Res. Geophys. J. t for = f m .Gohs Res. Geophys. J. ,w e tt variable state set we y, 0 + d 2 β ∼ pl within β 0.8 φ nta segment. that on ) f sfutguebulk gouge fault is ) m(al oe to core) (fault mm 1 = V t V p , pl t ξ 5 max ≈ p ece.Geophys. Geochem. < u oaunit a to due ) × d ≈ 10 V = 7175–7194 119, 10 10 sacriterion a as 9 max m With mm. 1 φ −4 3 B11102 117, p t for φ, 3 p o10 to 0 χ < η because s cosa across stheir as MPa = t when ) 10 p Nature k 10 ≈ 6 (x −3 φ V −1 −3 10 V pl − = φ , , , 1 .Smesn .Esot,C aoe nuneo iaac ntefitoa consti- frictional the on dilatancy of saturated Influence fluid Marone, C. of Elsworth, dilatancy D. Samuelson, Shear-induced J. Marone, 31. Anza C. the Elsworth, on D. depths Samuelson, locking geodetic J. and 30. seismicity Reconciling faults. Fialko, creeping Y. along Jiang, microearthquakes of J. Streaks Got, 29. L. J. Gillard, D. Rubin, from M. A. patterns 28. earthquake inferring Towards Bakun hydrothermal H. Lapusta, W. under N. gouge 27. Avouac, gabbro of J. sliding Kaneko, Frictional Y. Yao, 26. W. Wang, Z. He, C. constitutive laws. 25. friction and variable state results and instability Experimental Slip Ruina, 1. L. friction A. rock 24. of Modeling Dieterich, H. J. 23. 2 .J i,J .MGie .D en rcinlbhvo foenctasomfut and faults transform oceanic of behavior Frictional Behn, mechanism D. a M. McGuire, as J. strengthening J. Dilatant Liu, J. Rice, Y. R. J. 22. Bradley, M. Rubin, M. transients. A. deformation Segall, aseismic P. on dilatancy fluid-infiltrated 21. gouge fault a of of Role Rubin, instability M. slip A. Liu, by Y. and regulated compaction, 20. materials Dilatancy, Rice, granular R. of J. shear Segall, episodic P. Slow 19. Iverson, R. modeling N. constitutive Moore, and behavior L. P. Frictional Scholz, 18. H. C. Raleigh, B. C. swarms. earthquake Marone, of C. geology The 17. McGuire, J. J. Pacific East 16. the on systems transform of Segmentation Smith, K. D. Lin, J. Gregg, M. P. 15. eiwr o hi huhflcmet hthle oipoethe and Engineer- RGPIN/418338-2012 OCE-18-33279. improve and and Grants OCE-10-61203 to Sciences Discovery Grants NSF Canada Natural helped and of by RGPIN-2018-05389; that Council external supported Research anonymous was comments ing study two thoughtful This and their manuscript. Diggles, for Mike reviewers Detweiler, Shane Hickman, ACKNOWLEDGMENTS. at Framework Science opigcefiinsrahln-emsed-tt aus(22). S3 values Fig. steady-state long-term typically reach cycles, coefficients weak- earthquake coupling (velocity OTF zone multiple seismogenic the over in ening) accumulated slip total the to phases χ 2 Fig. in as parameters tancy 10 unie stasetsi (χ slip transient as quantified uin o esooy(RS aacne,ntokcd ZD, code network Insti- center, Research data Incorporated (IRIS) fdsn.org/networks/detail/ZD the Seismology at for archived tutions are experiment seismometer strongly not are Availability. Data positions of front range migration the the S5). in Fig. and thresholds Appendix , tested (SI fraction affected are Several 0.85 velocity snapshot. maximum to that contour the 0.6 of of the 0.8) fault on (e.g., the fraction positions a on western-most is velocity and respec- slip each eastern- are where migration For line the slip scales. as bilateral the temporal defined of and fronts tively west spatial 106 the and same to east longitude comparison the the snapshot, direct and at a migration 339.8 for km, swarm day 105 Gofar is at degree set longitudinal one is assuming with fronts diffusional modeled modeled events transient 1.2 to aseismic 0.7 of fronts tion Migration. Transient Aseismic period. loading interseismic the during creep slow extremely infiatprini eesdwhen because released dilatancy, strong is with not portion segments do three general significant slip the of a in in modes 1 is two to the up slip but add budget, necessarily Aseismic slip intersections. seismic the ridge–transform to complementary two the dilatancy of strong each under effect locking clear 5 a see can We strong, to 5)×10 weak from varies dilatancy of effect s × , 0.3 uiebhvo fastrtdfutzn ne ait fdang conditions. drainage of variety a under zone fault Res. saturated Geophys. a of behavior tutive theory. and Experiment faults: Fault. Jacinto San the of section Nature earthquake. Parkfield coupling. interseismic of observations geodetic conditions. (1983). 10370 equations. t nuneo atqaecharacteristics. earthquake on influence its events. slip slow for Res. Geophys. J. fault. strengthening. dilatant gouge. fault simulated of faults. transform oceanic fast-slipping at Geology processes earthquake for Implications Rise: χ 10 V a −4 and , −4 pl hw h ln-tiedsrbto of distribution along-strike the shows r lotse,adteoealsi ugtdpnec ndila- on dependence budget slip overall the and tested, also are , .Gohs Res. Geophys. J. h atrsoni e,i h central the in red, in shown latter the , 3–4 (1999). 635–641 400, lgtga odr ie) ihacntn ooiyvalue porosity constant a with lines), dark to gray (light × 8–9 (2006). 289–292 34, χ 10 .Gohs Res. Geophys. J. Tectonophys. t mlctosfrpeito n aadassmn rmte2004 the from assessment hazard and prediction for Implications al., et −4 146(2011). B10406 116, r l acltda h ai fsi eesdi respective in released slip of ratio the as calculated all are 144(2010). B10414 115, esi aeomdt rmte20 oa ca bottom ocean Gofar 2008 the from data waveform Seismic attlo ih iuaincss.Teoii fthese of origin The cases). simulation eight of total (a .Gohs Res. Geophys. J. 25–27 (1995). 22155–22171 100, Nature https://osf.io/rw9bq/. PNAS Geology etakJa obr,Rt ars Steve Harris, Ruth Gomberg, Joan thank We 5–6 (2007). 353–362 445, .Gohs Res. Geophys. J. 1126 (1979). 2161–2168 84, C 6–7 (2005). 969–974 437, aapout a eacse nOpen on accessed be can products Data 2007/. .Gohs Res. Geophys. J. | ooe oi ie nFg hwtemigra- the show 4 Fig. in lines solid Colored and epy.Rs Lett. Res. Geophys. 4–4 (2002). 843–846 30, a 2 2020 12, May t .Lwrctf eoiis 10 velocities, cutoff Lower ). 135(2010). B12305 115, D and 0772 (1990). 7007–7025 95, V .Gohs Res. Geophys. J. max i.S3 Fig. Appendix, SI 144(2009). B12404 114, a.Geosci. Nat. < χ | a.Geosci. Nat. s 10 , 0631,7 (2016). 10,663–10,671 43, = o.117 vol. χ 0.8 a 02 .,05 .,07 ,2, 1, 0.7, 0.6, 0.5, 0.3, (0.2, 0k and km ∼10 and , V .Gohs Res. Geophys. J. pl 035(2012). B04315 117, 6–6 (2010). 363–369 3, 0cce hnthe when cycles ∼10 hc consfor accounts which , χ 28 (2019). 82–87 12, | φ t lprto sthe as ratios slip = = o 19 no. sntaffected. not is 5and 0.05 2 https://www. IAppendix, SI × 5k at km ∼15 0. 10 5 φ 10359– 88, V | .262 −4 = pl 10193 0.05. V and and max ◦ = S, J.

EARTH, ATMOSPHERIC, AND PLANETARY SCIENCES 32. R. Lohman, J. McGuire, Earthquake swarms driven by aseismic creep in the Salton 39. S. Wei et al., The 2012 Brawley swarm triggered by injection-induced aseismic slip. Trough, California. J. Geophys. Res. 112, 1–10 (2007). Earth Planet Sci. Lett. 422, 115–125 (2015). 33. H. Hirose, T. Matsuzawa, T. Kimura, H. Kimura, The Boso slow slip events in 2007 and 40. A. Llenos, J. McGuire, Y. Ogata, Modeling seismic swarms triggered by aseismic 2011 as a driving process for the accompanying earthquake swarm. Geophys. Res. transients. Earth Planet Sci. Lett. 281, 59–69 (2009). Lett. 41, 2778–2785 (2014). 41. X. Chen, P. Shearer, R. Abercrombie, Spatial migration of earthquakes within seis- 34. H. Dragert, K. Wang, T. S. James, A silent slip event on the deeper Cascadia subduction mic clusters in Southern California: Evidence for ﬂuid diffusion. J. Geophys. Res. 117, interface. Science 292, 1525–1528 (2001). B04301 (2012). 35. K. Obara, H. Hirose, F. Yamamizu, K. Kasahara, Episodic slow slip events accompanied 42. Y. Ito et al., Episodic slow slip events in the Japan subduction zone before the 2011 by non-volcanic tremors in southwest Japan subduction zone. Geophys. Res. Lett. 31, Tohoku-Oki earthquake. Tectonophys. 600, 14–26 (2013). L23602 (2004). 43. E. Brodsky, T. Lay, Recognizing foreshocks from the 1 April 2014 Chile earthquake. 36. L. A. Reinen, J. D. Weeks, T. E. Tullis, The frictional behavior of lizardite and antigorite Science 344, 700–702 (2014). serpentinites - experiments, constitutive models, and implications for natural faults. 44. A. M. Rubin, Episodic slow slip events and rate-and-state friction. J. Geophys. Res. 113, Pure. Appl. Geophys. 143, 317–358 (1994). B11414 (2008). 37. E. Roland, J. J. McGuire, Earthquake swarms on transform faults. Geophys. J. Int. 178, 45. J. R. Rice, Spatiotemporal complexity of slip on a fault. J. Geophys. Res. 98, 9885–9907 1677–1690 (2009). (1993). 38. S. Holtkamp, M. Brudzinski, Megathrust earthquake swarms indicate frictional 46. K. Mizoguchi, T. Hirose, T. Shimamoto, E. Fukuyama, Internal structure and per- changes which delimit large earthquake ruptures. Earth Planet Sci. Lett. 390, 234–243 meability of the Nojima fault, southwest Japan. J. Struct. Geol. 30, 513–524 (2014). (2008).

10194 | www.pnas.org/cgi/doi/10.1073/pnas.1913625117 Liu et al. Downloaded by guest on October 1, 2021