Time-variable Deformation in the New Madrid Seismic Zone

(why there, why now?)

Eric Calais, Andy Freed, Purdue UniversitySeth Stein, Northwestern UniversityRoy van Arsdale, University of Memphis

Research supported by the USGS, NEHRP program

American Geophysical Union 2009 Joint Assembly

We thank CERI (Memphis) and NGS/CORS for making their GPS data publicly available

• Large earthquakes happen in plate interiors→ No long-term topography anomaly associated→ Why do they occur?→ How often do they repeat – and for how long?

• New Madrid Seismic Zone key to this issue→ Sequence of 3 M7-7.5 events in 1811-1812→ Well instrumented region (seismic, geodetic, paleoseismo)→ Vulnerability high→ Test-bed for steady-state elastic rebound model in plate interiors

M>6 - NEIC catalog, historical

+ recorded

• Prediction at linear rate is 0.3 mm/yr at t = 2009

… measurement at 2009 = not significantly different from zero, upper bound = 0.2 mm/yr

?

(1) Liu et al., 1992(2) Snay et al., 1994(3) Argus and Gordon, 1996; Dixon et al., 1996(4) Weber et al., 1998(5) Newman et al., 1999(6) Gan and Prescott, 2001(7) Sella et al., 2002(8) Marquez-Azua and DeMets, 2003(9) Smalley et al., 2005(10) Calais et al., 2005(11) Calais et al., 2006

• Glacial Isostatic Adjustment signal stands out in the vertical and horizontal GPS

• Forebuldge area:− Strain rates up to ~2x10-9 yr-1

− Little seismicity overall (except St Lawrence and NE US)

• South of forebuldge:− No significant strain (WRMS of residual

velocities = 0.5 mm/yr)− Largest earthquakes

Calais et al., JGR 2006

Yellow arrows = filtered GPS velocities w.r.t. stable North America

Background colors = interpolated vertical ratesBlack arrows = interpolated principal strains

New MadridNew Madrid

Current plate-widedeformation

Seismicity – NEIC catalog

0.5 +- 1.4 mm/yr 0.0 +- 0.2 mm/yr0.8 +- 0.9 mm/yr

New Madrid:

• Residual velocities < 0.2 mm/yr• Strain rate < 1.3x10-9 yr-1

• uncertainties and residual velocities have decreased by at least a factor of 2 at all sites

• Sites with the worse quality position time series such as RLAP also have the largest residuals

(more details in Calais and Stein, Science, 2009)

Van Arsdale et al., 2000

• M~7 events in the NMSZ happen every ~500 years over the past 3,000 years

• NMSZ deforms at < 0.2 mm/yr => one M7 every 10,000 years

Steady-state elastic rebound model fails

Time-variable deformation

… consistent with NMSZ activated ~10,000 years ago

Tuttle et al. (2002)

Recently activated fault zone with ~no far-field displacement loading, no near-field strain accumulation, but sequences of large earthquakes that are temporally clustered and migrating…

•How was the NMSZ activated?−GIA – requires ad hoc weak mantle (Grollimund and Zoback,

2001)

−Sinking of high-density crustal body – requires strain rates ~10-7 yr-1 (Pollitz et al., 2001)

−Ad hoc weak crust – requires strain rates ~10-8 yr-1 (Kenner and Segall, 2000)

•How are sequences of large earthquakes maintained in the absence of far-field loading (and for how long)? viscoelastic stress transfer (Kenner and Segall, 2000; Pollitz et al., 2001)

•Why do earthquakes migrate? stress transfer between neighboring faults (Li et al., 2009; Stein et al., this meeting)

Why New Madrid? What makes it special?

McKenna et al., 2007

- Heat flow? No anomaly.

- Paleorift? Pre-existing faults, but other paleorifts show no seismicity.- Strain anomaly? None detected geodetically so far, no topography.- Stress anomaly? Secular stress anomaly would have triggered

earthquakes long ago.- Sudden weakening? No clear mechanism why…

Forte et al., 2007

• NMSZ coincides with upper Mississippi Valley (Eastern Lowlands)

• Eocene overlain by early Pliocene “Upland Complex Gravels” and Pliocene loess

• Up to 100 m of erosion in the past 4 Ma (since UG) in 2 events

Modified from VanArsdale et al., 2007

• Pliocene incision:− 4-1.8 My− Sea level change− Slow

• Holocene incision:− 18,000-10,000 BP− Confluence of Mississippi

and Ohio Rivers− Rapid

Model ingredients:• Erosion (= unloading) history from

geologic mapping (VanArsdale et al, 2007)

• Flexural response with time and amplitude characteristics dependent on lithosphere/mantle rheology− Viscoelastic lithosphere/mantle− Depth-dependent elastic properties− Temperature-dependent viscosity

(CEUS geotherm from Mc Kenna et al., 2007)

− Laboratory flow law for wet olivine and =10-13 yr-1

• Boundary conditions = zero displacements

• Reelfoot rift faults at failure equilibrium (cf. Townend and Zoback, 2000)

Questions:• Are flexural stresses sufficient to

trigger earthquakes (~1 MPa)• … while matching geodetic

observations (strain rate < 10-9/yr)

Horizontal (clamping) stress changes

load

0

60 km

Horizontal strain rateload

0

60 km

Uplift rate

load

0

60 km

Conclusions

• Observations:– Current horizontal deformation in stable North America less than 0.5

mm/yr (plate wide) and 0.2 mm/yr (New Madrid)

– NMSZ activated in the Holocene.

– NMSZ experienced 4 M~7 events in 2,000 years.

• What makes the NMSZ special? Its Holocene geological history.• Proposed scenario:

– Unloading due to climate-related erosion in upper Mississippi valley causes (unclamping) flexural stresses sufficient to activate the NMSZ

– Once activated:• Seismicity can be maintained via viscoelastic relaxation processes until stresses are fully

dissipated (Kenner and Segall, 2000)

• Seismicity can migrate in time and space due to stress transfer between neighboring faults and limited far-field loading (Li et al., 2008)

– Open questions: role of rheology, how long can seismicity be maintained, why large earthquakes, spatial scale of strain accumulation?