Dealing with paradoxes in subduction zone geodynamics Kelin Wang1,2 1Pacific Geoscience Centre, Geological Survey of Canada 2School of Earth and Ocean Sciences, University of Victoria Acknowledgements: Yan Hu – deformation modeling (PhD work) Ikuko Wada – thermal modeling (PhD work) John He – computer programming

Wada and Wang, 2009, G3

Max. Depth of a Low-Velocity Layer Deeper basalt-eclogite transformation and peak crustal dehydration Slab thermal parameter (102 km) = Slab age × Descent rate (Fukao et al., 1983;Cassidy and Ellis, 1993; Bostock et al., 2002; Hori et al, 1985; Hori, 1990; Ohkura, 2000; Yuan et al., 2000; Bock et al., 2000; Abers, 2006; Rondenay et al., 2008; Matsuzawa et al., 1986; Kawakatsu and Watada, 2007)

Depth Range of Intraslab Earthquakes Dehydration embrittlement at deeper depths Slab thermal parameter (102 km) = Slab age × Descent rate (Inferred from earthquakes located by Engdahl et al. 1998 and local networks)

Intensity of Arc Volcanism More magma production Slab thermal parameter (102 km) = Slab age × Descent rate (Crisp, 1984; White et al., 2006)

Survival depth of basaltic crust (blue diamond) and Warm Cold Survival depth of basaltic crust (blue diamond) and depth range of intraslab earthquakes (purple lines) Eruption rate of arc volcanoes (White et al., 2006) Depth of slab beneath volcanic arc Colour: different publications Wada and Wang, 2009, G3

Paradox 1 Subduction zones exhibit great (thermally controlled) diversity in petrologic, seismic, and volcanic processes, but they share a rather uniform slab-arc configuration.

Low seismic attenuation Low Vp/Vs Serpentinization Stagnant Cold Forearc Hot Arc, Back Arc 70 ~ 80 km ~ 100 km Low seismic attenuation Low Vp/Vs Serpentinization Stagnant High attenuation High Vp/Vs Melting Vigorous wedge flow

Northern Cascadia (Currie et al. 2004, EPSL)

inflow outflow Oceanic geotherm (plate cooling model) Landward Temperature- and stress-dependent mantle wedge rheology Depth Mantle adiabat outflow Temperature

Heat Flow Measurements

Heat flow transect across the Cascadia subduction zone probe BSR Offshore well Land borehole ODP hole

Comparison with thermal model results

Two primary constraints: Blue: Basaltic crust Purple: Serpentine stability in slab or mantle wedge Preferred Cascadia model Decoupling to ~ 70 - 80 km depth Two primary constraints: Surface heat flow (cold foreac) Mantle temperature beneath arc > 1200C (hot arc)

Fluid content in the subducting slab Crust (wet basalt) Mantle warm cold wt% bound H2O Phase diagram from Hacker et al. (2004) Reactions from Schmidt and Poli (1998) Wet solidus: (1) Schmidt and Poli (1998), (2) Grove et al. (2003)

End-member warm-slab and cold-slab subduction zones N Cascadia NE Japan Blue: Basaltic crust Purple: Serpentine stability Basalt to eclogite ~ 40-50 km depth Feeble arc volcanism Serpentinized mantle wedge corner Intraslab earthquakes to ~90 km depth Basalt to eclogite ~ 100-140 km Active arc volcanism High-velocity wedge corner Earthquakes to hundreds of km Kirby et al., 1996; Wada and Wang, 2009; Syracuse et al., 2010 ; van Keken et al., 2011

Assuming decoulping to 75 km Wada and Wang, 2009, G3

Survival depth of basaltic oceanic crust (blue) and Warm Cold Survival depth of basaltic oceanic crust (blue) and depth range of intraslab earthquakes (purple) Model-predicted peak dehydration depth (blue) and antigorite stability in subducting slab (purple) Wada and Wang, 2009, G3

Paradox 1: Subduction zones exhibit great (thermally controlled) diversity in petrologic, seismic, and volcanic processes, but they share a rather uniform slab-arc configuration. Reconciliation: Common depth of decoupling between the slab and the mantle wedge

Weakening of slab-mantle wedge interface Weak hydrous minerals: (wet) serpentine, talc, brucite, chlorite e.g. frictional coefficient  of wet talc ~0.2 Elevated fluid pressure: if  = 0.2, Pf /Plith = 90%,  = 0.02

? 1 1

Northeast Japan 1 3

2 or 3 1 Hellenic Arc Southeast Mexico Quaternary faults (Angelier et al., 1982) and earthquake focal mechanisms (Benetatos et al., 2004) 1 2 or 3 Southeast Mexico

Northern Cascadia 1 2 or 3

Summary of Stress Indicators

Paradox 2 Subduction zones accommodate plate convergence, but few forearcs are under margin-normal compression.

Contours of maximum shear stress Mantle wedge rheology: Far-field force Contours of maximum shear stress Mantle wedge rheology: Dislocation creep Effective viscosity:

?   n Summary of Stress Indicators Force Balance Model Assuming V = H, Lamb (2006) obtained   0.03 for most subduction zones   0.05 ?

Red: Stress constrained by stress indicators I compiled. Blue: Megathrust stress determined by Lamb (2006) assuming V = H. Thermal models have been developed for most sites with   0.03 for frictional heating along megathrust.

Do Chilean-type subduction zones have a strong fault? Modeling Results for Peru-Chile Lamb (2007):   0.095 assuming V = H Richardson and Coblentz (1994): H=25 MPa (  0.06) recognizing V > H Sobolev and Babeyko (2005):  = 0.015  0.05 orogeny model

Paradox 2: Subduction zones accommodate plate convergence, but few forearcs are under margin-normal compression. Explanation: Plate interface too weak to overcome gravitational tension in the forearc.

small earthquakes in upper plate Summary of Stresses in Cascadia forearc small earthquakes in upper plate Wang, 2000, Tectonophysics

A 100-km line becomes shorter by 2 cm each year Geodetic Strain Rates

small earthquakes in upper plate Geodetic Strain Rates Forearc Stresses Wang, 2000, Tectonophysics

Nankai Forearc Stresses and geodetic strain rates are similar to Cascadia Wang, 2000, Tectonophysics

Paradox 3 At some forearcs, maximum compression is margin-parallel, but fastest geodetic shortening is roughly margin-normal.

If deformation is elastic, it only reflects stress changes and has nothing to do with absolute stress. Cascadia geodetic shortening reflects stress increase due to interseismic locking of the plate interface. Geodetic Strain Rates

A Stretched Elastic Band Time 1: Tension Contraction Time 2: Less tension

If deformation is elastic, it only reflects stress changes. Cascadia geodetic shortening reflects stress increase due to interseismic locking of the plate interface. Great earthquake cycles cause small perturbations to forearc stress. Geodetic Strain Rates

(Probability from inversion) >20% peak slip Entire fault Static stress drop (Probability from inversion) Areas with >10% peak slip If deformation is elastic, it only reflects stress changes. Cascadia geodetic shortening reflects stress increase due to interseismic locking of the plate interface. Tohoku earthquake Mw = 9 March 11, 2011 Great earthquake cycles cause small perturbations to forearc stress. Simons et al., 2011

Margin-normal stress perturbation Margin-parallel compression

Margin-normal stress perturbation Margin-parallel compression

Paradox 3: At some forearcs, maximum compression is margin-parallel, but fastest geodetic shortening is roughly margin-normal. Explanation: The geodetic shortening only reflects small stress changes in earthquake cycles.

Cascadia: All sites move landward Wells and Simpson (2001) Wang, 2007, SEIZE volume

Alaska and Chile: Opposing motion of coastal and inland sites Freymueller et al. (2009) Wang et al. (2007, G3)

Paradox 4 Interseismic locking of subduction fault causes landward motion of the upper plate, but some areas show seaward motion.

Japan and Sumatra: All sites move seaward 3.5 months after M=9 quake A few years after M=9.2 quake http://www.gsi.go.jp/cais/topic110314-index.html Grijalva et al (2009)

Based on Wang, 2007, SEIZE volume Coast line Inter-seismic 2 (Cascadia) Inter-seismic 1 (Alaska, Chile) Post-seismic (Japan, Sumatra) Co-seismic Coast line Based on Wang, 2007, SEIZE volume

Locking Rupture Afterslip Stress relaxation Stress relaxation Characteristic timescales: Afterslip – months to a few years Viscoelastic relaxation (transient) – a few years Viscoelastic relaxation (steady-state) – a few decades Locking – (centuries) length of the earthquake cycle

A couple of years About four decades Three centuries

 M K TK = 10K/= 3 yr Central part of Sumatra mesh Hu, 2011, PhD thesis M K  TM = 10M/ = 60 yr TK = 10K/= 3 yr Central part of Sumatra mesh

A couple of years About four decades Three centuries Wang et al., in prep.

Deformation Following the 1700 Cascadia Earthquake 2 yr after EQ (like Japan, Sumatra) 40 yr after EQ (like Chile, Alaska) Present Hu, 2011, PhD thesis

1995 Antofagasta earthquake, N. Chile (Mw = 8.0) 1993-95 Displacements (dominated by co-seismic) 1996-97 Velocities (2 years after earthquake) Data from Klotz et al. (1999) and Khazaradze and Klotz (2003)

Paradox 4: Interseismic locking of subduction fault causes landward motion of the upper plate, but some areas show seaward motion. Explanation: The seaward motion is the result of afterslip and viscoelastic mantle relaxation. It will diminish with time.

Paradox 5: Mountain building at a subduction zone Paradox 6: Episodic tremor and slip Paradox 7: Strong asperities of weak faults Paradox 8: … … … … Paradox 1000: … … To be continued … …

… …

In model: Coupling stress represented by ’ and h Moho Layer viscosity ’ Thickness h In Earth: Interface and wedge strengths controlled by petrology and fluid In model: Coupling stress represented by ’ and h

Wang and He, 1999, JGR