Perspectives of Japan

1
Coseismic Folding and Stress Transfer in the
1811/12 New Madrid Earthquakes Karl Mueller
Department of Geological Sciences, University
of Colorado, Boulder COReconstruction of Earths
tectonic plates argues that lithosphere in
continental interiors moves rigidly and should
not otherwise deform. Yet earthquakes in the
central United States in the New Madrid seismic
zone suggest that some strain is accommodated in
the interior of the North American plate. Trench
excavations, seismic reflection profiling and
recording of micro earthquakes suggests that
strain in New Madrid is accommodated by slip
across a large restraining bend on a linked
right-lateral strike slip and thrust fault
system. Four large earthquakes occurred in the
zone during a two month period in1811-1812 AD
that ranged in magnitude from Mw 6.8 to Mw7.5
and were felt as far as 1500 km away on the
eastern seaboard of the United States. Previous
earthquake sequences in the zone occurred in 1450
AD and 900 AD based on studies of liquefaction
deposits in the Mississippi River floodplain.
Surface deformation in the zone is best expressed
by coseismic folding of late Holocene fluvial
sediments along the Reelfoot scarp, which
comprises the forelimb of a fault-propagation
fold above a blind thrust fault. Reconstruction
of folded late Holocene sediments in the last
three earthquake cycles in New Madrid suggest the
Reelfoot thrust has slipped at a rate of 4mm/yr
and that total slip on the fault is less than 100
meters. If the current slip rate is assumed to
be constant since the inception of movement on
the zone, this implies that the age of modern New
Madrid fault system must be less than about 25ka.
New work based on modeling of Coulomb elastic
stress transfer, historical accounts of
liquefaction and surface rupture and monitoring
of microseismicity suggests that the third
earthquake in the 1812 AD sequence actually
happened 200km northeast of the main New Madrid
zone as a remotely triggered event. Models of
Coulomb stress change caused by the other three
earthquakes suggest that earthquakes on adjacent
faults triggered one another with the first event
occurring on the strike-slip Cottonwood Grove
fault, the second on a much smaller strike slip
or thrust fault and the last and largest on the
Reelfoot thrust. The origin of the stress field
that causes earthquakes in the New Madrid seismic
zone is poorly understood, however the stress
field located further north may be produced by
crustal rebound caused by melting of a
continental ice sheet at the end of the last
glacial period in the late Pleistocene.
2
Location of the intraplate New Madrid seismic
zone, central USA Red Earthquakes White
Floodplain Tan Uplands
3
Digital elevation model New Madrid
Lake County Uplift Mississippi
River Older meanders Reefoot Lake
4
Mississippi River 1-3 km wide, largest river in
USA
1.5 km
5
Reelfoot Lake, formed by coseismic folding
6
New Madrid in 1811 - The American Frontier
7
Past E-quakes in New Madrid Tuttle and Schweig
(2003) NEHRP Report
8
Liquefaction Evidence for Late Holocene
Earthquakes

• Last four clustered cycles have 360-600 year
  recurrence
• Prior cycle has longer recurrence 2600 years

9
DEM of New Madrid
Floodplain less than 2300 years old near
uplift Uplands made of glacial loess Meander
belt only 7.5 ka
10
River Floodplain
Lake County Uplift Reelfoot Scarp Ridgely
Ridge Mississippi River Past
meanders Reefoot Lake
Mississippi
Reelfoot Scarp
Lake
Ridgely
Ridge
Cottonwood Grove Fault
11
Uplifted remnants smaller than original structure
12
Oblique DEM of New Madrid
13

• Meander chronology
• 2.3 Ka record
• resets topography

14
Northern segment
Central segment
Southern segment

• Morphology and Relief of Reelfoot Scarp
• Hundreds of meters wide
• Northern segment records 5m of topographic
  relief
• Central segment mostly covered by lake waters
• Southern segment decreases in elevation
• Scarp terminates in Ridgely Ridge, not present
  further south

15

• Trench excavation across Reelfoot Scarp
• See Mueller et al., (1999), Champion et al,
  (2001)
• 4mm/yr slip rate on Reelfoot thrust, implies
  1.0mm/yr
• horizontal contraction across zone
  (matches GPS)
• 3 smaller scarps suggest strain distributed
  unevenly
• 9m of uplift in 2300 years (Guccione et al, 2002)
• Secondary extensional strain, no flexural slip
  features
• Pervasive liquefaction features, some very large

16

• Mini Sosie
• seismic profile
• of Reelfoot Scarp
• Line LDC-2 from
• Sexton Jones (1986)
• fault propagation fold
• flattening of reflectors
• fault offset at 0.6sec TWTT
• no growth strata

17

• Inverse Trishear
• Model for Reelfoot
• fault-propagation fold (LDC-2)
• restores strata
• defines fault dip
• propagation to slip 9
• reactivated fault

18

• Forward Model
• Reelfoot fault-
• propagation fold
• (profile LDC-2)
• Deep fold
• good match
• Shallow fold
• model over-
• predicts slip
• (old growth?)

19
Obion River Mini Sosie Profile
20

• Inverse trishear
• model for Reelfoot
• fault-propagation
• fold (Obion profile)
• better restoration
• steeper fault dip
• propagation to
• slip 0.9
• reactivated fault
• not active now?

21

• Forward Model Reelfoot fault-propagation fold
  (profile SRL-3)
• Outstanding model fit, less slip, very steep
  fault, but inactive?

22
Seismicity maps surface of Reel- foot thrust
(Mueller Pujol, 2002) Reelfoot thrust is
well defined Northern Arms Are nearly
vertical zones of seismicity (ie not faults?)
23
Central segment Consistent fault
geometry No shallow seismicity
24
Intersection of Reelfoot thrust and dextral
Ridgely Cottonwood Grove faults obscure thrust
geometry
25
Southern thrust segment is better defined and
shows a steeper thrust segment (but it may not
be active given a lack of surface deformation)
26

• Structure contour map
• of Reelfoot thrust
• from microseismicity
• defines area of fault
• for 1812 Mw estimates
• (much lower than earlier Mw 8 -8.3 estimates)
• not all seismicity relates to fault rupture in
  1812

27
(No Transcript)
28
Moment released Jan 23, 1812 thrust
earthquake MouDA (Hanks and Wyss, 1972) U
crustal ridigity (3.5 x 1011 dyne cm-2 (high in
midplate) D fault displacement (1.7-1.9m) A
fault area (1300 km2) Mo moment (7.6 to 8.6 x
1026 dyne-cm Mw 0.67log Mo - 10.7 (Hanks and
Kanamori, 1979) Mw 7.2 to 7.3 - Jan 23, 1812
thrust earthquake Much lower than previous
estimates of Johnston (8.0-8.3)!
29
Earthquake Scenario I From Coulomb stresses NM1
16 Dec 1811 5m dextral slip, Mw 7.3 Cottonwood
Grove (Source) Loads NM1A 16 Dec 1811 1m
dextral slip, Mw 7.0 Cottonwood Grove
(Receiver) Does not enhance stresses for NM2
23 Jan 1812 1m dextral slip. Mw 7.0 Northeast Arm
(Receiver)
30
Earthquake Scenario II NM1 16 Dec 1811 5m
dextral slip, Mw 7.3 Cottonwood Grove
(Source) NM1A 16 Dec 1811 1m dextral slip, Mw
7.0 Cottonwood Grove (Source) Enhances stresses
for NM3 27 Feb 1812 5m reverse slip, Mw
7.5 Reelfoot Blind Thrust (Receiver)
31
Earthquake Scenario III (preferred) NM1 16 Dec
1811 5m dextral slip, Mw 7.3 Cottonwood Grove
(Source) NM1A 16 Dec 1811 1m reverse slip. Mw
7.0 Reelfoot Blind Thrust (Source) Enhances
stresses for NM3 27 Feb 1812 5m reverse slip,
Mw 7.5 Reelfoot Blind Thrust (Receiver)
32
Preferred Earthquake Scenario in
1811/1812 Microseismicity corresponds well to
predicted areas of increased Coulomb stresses
Only two faults actually ruptured in
1811-1812, the right lateral Cottonwood fault
and the Reelfoot Blind thrust. Other arms
of microseismicity are actually off fault lobes
of increased stress. Interesting note
After- shocks in midplate settings can last for
centuries based on rate and state friction laws.
33
Analysis of NM2 Using the method of Bakun
and Wentworth, BSSA (1997) Hough determines
location of NM2 to be in southern Illinois, 200
km from New Madrid seismic zone Location
corresponds to newly discovered written record
of 2km- long surface rupture and
extensive liquefaction deposits (sand blows) in
Wabash Valley, a region of high microseismicity R
esults consistent with Coulomb stress modeling
that suggests NM2 was not enhanced by prior
events (NM1 and NM1A)
34
One historic account provides an intriguing
suggestion of a possible surface rupture some 220
km from New Madrid. This account, by Mr. Yearby
Land, described a a big crack that was made in
the ground, with two feet of vertical
displacement to the south (23). Even in 1858 the
feature (38.07N, 88.11W) could be traced for a
reported distance of two miles. Near this crack
Mr. Land stated that piles and piles of pure,
snow white sand were heaved up, including some
as big as several wagon loads. Field
reconnaissance has verified many of the details
of the Land account and confirmed evidence of
sand blows on the surface of the field where they
were reported. Mueller, Hough and Bilham,
Nature, in press
35
(No Transcript)
36
Origin of stresses north of New Madrid is
probably affected by melting of late Pleistocene
continental ice sheet and resulting crustal
rebound.
37
Map of Eastern Canada showing distribution of
M3.0 and greater earthquakes from 1980 to 1990
and M6.0 and greater earthquakes since 1663.
Symbols are defined as follows BU - Boothia
Uplift BA - Bell Arch BB - Baffin Bay LS -
Labrador Sea GB - Grand Banks SLV - St.
Lawrence Valley OBG - Ottawa Bonnechere Graben.
This map shows that most of the seismic activity
in Eastern Canada is along the following three
tectonic trends or pre weakened zones i) the
Baffin Bay-Grand Banks portion of the Mesozoic
rifted margin, ii) the reactivated Paleozoic
structures along the Boothia Uplift-Bell Arch and
iii) the Paleozoic rifts along the St. Lawrence
Valley-Ottawa Bonnechere Graben. Data taken
from Canadian National Earthquake Database
Although Eastern Canada (east of the Rocky
Mountains) is on a stable craton and is far away
from any active plate boundaries, it still
experiences earthquakes with magnitudes up to M7
(See Fig.1). A noteworthy example is the 1989
Ungava earthquake of magnitude (Ms) 6.3 that has
ruptured the Earth's surface.
38
Map of Eastern Canada and peripheral area
showing (i) orientation of the contemporary
regional stress field (bold inward-pointing
arrows) on land and offshore Labrador. (ii) the
orientation of the stress in an earlier time as
indicated by postglacial faults (red
inward-pointing arrows). Modified from Adams GSC
Open File 3122,1995. Earthquakes are closely
related to the stress field. The orientation of
the contemporary stress field on land is fairly
uniform with SHmax aligned in the NE direction,
and can be explained by ridge-push forces at the
Mid-Atlantic. Thus, rebound stress has little
effect on contemporary stress orientations on
land. Postglacial faults, however, indicate that
the SHmax orientation of the paleo-stress field
9,000 years ago was close to the NW, consistent
with the direction of ice retreat. Thus stress
orientation has rotated significantly during the
last 9,000 years. This stress rotation is due to
changes of the dominant stress component from the
transient rebound stress 9,000 years ago to
tectonic stress at the present.
39
Contour plot of Ice thickness (in Meters) at 18
KBP
1) calculate the spatio-temporal evolution of the
total stress field in E. Canada. The total stress
consists of rebound stress, tectonic stress and
overburden stress. Here, rebound stress refers to
the stress induced in the mantle and lithosphere
during the loading and unloading of the late
Pleistocene ice sheets. At seismogenic depth,
rebound stress is mainly due to lithospheric
flexure, however creep in the mantle modifies the
thickness of the "effective lithosphere" and thus
rebound stress changes with time continuously.
The rebound stress due to the application of the
ICE3G load on a viscoelastic Earth is calculated
using the Finite Element Method. Visco-elastic
earth models with different viscosity profiles
have been considered. The orientation of the
tectonic maximum horizontal principal stress is
given by the World Stress Map Project, which
shows that the first order stress in E. Canada is
along N60E - consistent with ridge-push at the
mid-Atlantic Ridge.
40
2) relate total stress to earthquakes. According
to the Coulomb-Mohr Theory of Failure,
optimally-oriented pre-existing faults can be
reactivated when the Mohr circle (representing
the total stress field) touches the Line of
Failure. FSM measures fault stability and is the
shortest distance from Mohr's circle to the line
of failure. FSM will change when the state of
stress changes. Thus, for any spatial location,
the evolution of total stress gives the changes
in the Fault Stability Margin (dFSM) at time t,
where dFSM(t) FSM(t) - FSM(to) and to is the
initial time before the onset of deglaciation. A
positive dFSM means fault stability is promoted
while a negative dFSM means fault instability is
promoted. 3) determine the Mode of Failure. This
depends on which principal stress is closest to
the vertical. For example, if the maximum
principal stress is nearly vertical, then the
mode of failure is normal (see diagram above). On
the other hand, if the minimum principal stress
is vertical, then the mode of failure is
Thrusting. 4) determine the stress orientation
from the total stress field
Summary of Model Input/Output
41

• For earthquake studies, strain rate maps are
  usually useful for revealing the earthquake
  source. Thus, spatial-temporal variation of
  Strain Rate have been computed and plotted.
  Diagram on the left shows that the normal
  vertical strain rate (Ezz), which is related to
  the vertical uplift velocity due to postglacial
  rebound, is largest inside the former ice margin.
  However, this map shows little correlation with
  the observed spatial pattern of seismicity
  (Fig.1).

42
Next, let us explore the vertical shear strain
rate (Ezh), which is related to the shear stress
of postglacial rebound. They are shown on the
diagram on the left. Although there is good
correlation between the location of maximum shear
rate and the location of earthquakes (Fig.1)
along the Baffin Island-Labrador coast,
Southampton Island and the mouth of St. Lawrence,
little earthquake activities are observed around
Lake Superior-Lake Winnipeg, James Bay and
southern Greenland where peak/trough in shear
strain rate are predicted. This indicates that
shear stress is not the only factor controlling
earthquakes. Obviously, we need to take into
account fault stability margin (FSM) and its
changes (dFSM). Thus dFSM will be investigated in
the following
43
Spatio-temporal variation of dFSM in Eastern
Canada and peripheral area is shown at 4 time
steps.
dFSM at 12,000 year ago. Contours are in MPa.
dFSM at 18,000 year ago. Contours are in MPa
dFSM at 9,000 year ago. Contours are in MPa.
dFSM at the present time. Contours are in MPa.
Spatio-temporal variation of Fault Stability
(dFSM) in Eastern Canada and peripheral area is
shown at 12,000 years ago (Fig.4) and 9,000 years
ago (Fig.5). Contours are in MPa. Positive value
of dFSM (green, yellow, red) means that faults
are stabilized (Mohr circle moving away from
failure). Negative values of dFSM (light and dark
blue) means that fault instability/earthquakes
are promoted (Mohr circle towards failure). Fig.
4 shows that fault stability is promoted under
the ice load.
44

• Fig.6 Changes in Stress and dFSM at Other sites

Evolution of the predicted horizontal principal
stresses (Hmax, Hmin), the vertical principal
stress (PrinZ), dFSM and mode of failure at a
depth of 12.5 km for different sites are
predicted. Two typical examples are shown in
Fig.6. The total stress is the superposition of
rebound stress, tectonic stress and overburden
stress. Our model predicts that near Charlevoix
(Quebec), fault instability is promoted around
9500 BP and seismic activities should be maximum
around 9,000 BP. This predicted onset time agrees
with the observed onset time of 9,000 BP (red
arrow) in nearby Lac Temiscouata. The mode of
failure predicted is Thrust-Faulting - in
agreement with the observations. In Indiana, the
predicted onset time also agrees with the
observed timing of the nearby Wabash Valley
earthquake.
The effect of mantle viscosity has also been
investigated. The results shows that in general,
the effect of mantle viscosity on the onset time
is generally small, except for sites near the ice
margin. The effect of high viscosity lower mantle
on the onset time and the mode of failure is
usually small. This is demonstrated for a model
with 1E22 Pa-s lower mantle. However, for
Laurentia, mantle viscosity strongly affects the
answer to the question " Will earthquake
activities increase in the next few thousand
years?" This is an important question for the
location of nuclear waste repositories where
fault stability is required for the next few
thousand years.  
45
Fig.7 shows the predicted orientation and
magnitude of the horizontal principal stresses
(a) 9000 years ago and (b) at the present. This
is obtained by superposing a static tectonic
stress in the NE with differential magnitude of 5
MPa onto the time-dependent rebound stress due to
loading of the ICE3G model on a stratified,
viscoelastic Earth that has a uniform 1.E21 Pa-s
mantle. Fig.7 shows that for this Earth model
where rebound stress relaxes rapidly, significant
stress rotations during the last 9,000 years is
predicted. Moreover, the contemporary first order
stress field is predicted to have uniform
orientations. These are in agreement with the
observations (Fig.2). Ice models with improved
spatial resolutions are needed to make the
predicted paleostress orientations agree with
those observed in southeastern Canada. Stress
orientation can also be used to constrain
tectonic stress difference too!
46
For an Earth model that has a high 1.E22 Pa-s
viscosity lower mantle, rebound stress relax very
slowly. As a consequence, rebound stress remains
the dominant stress component and very little
stress rotations occur during the last 9,000
years. Also, very non-uniform stress orientations
are predicted for the present. These are contrary
to the observed data as summarized in Fig.2 Thus,
the viscosity of the mantle can be inferred from
the stress rotation data.
47
Conclusions 1) Tectonic events of the past
created pre-weakened zones where most current
earthquakes of Eastern Canada are located.
However, rebound stress has been and still is
able to trigger seismic and faulting activity
within these preweakened zones. 2) At
postglacial time, thrust faults formed in
response to rebound stresses that controlled the
orientation of the total stress field at that
time. However, since the end of deglaciation
9,000 years ago, these rebound stresses gradually
diminished until, at present, the plate tectonic
stresses appear to be the dominant forces
influencing the orientation of SHmax. Associated
with the decay of rebound stress is the rotation
of the orientation of the total stress field.
This is consistent with the observed stress
rotations during the last 9,000 years. 3)
Within the ice margin, seismic and faulting
activities are suppressed by glacial loading,
however, soon after deglaciation, thrust faulting
is activated in preweakened tectonic zones. Thus,
the current mode of failure and the thrust motion
of the postglacial faults can be explained.
Furthermore, if we assume that the strength of
rocks is strong (i.e. gt 1 MPa of initial Fault
Stability Margin) everywhere except for the
preweakened tectonic zones where faults are
initially close to failure, then the observed
current spatial distribution of earthquakes can
also be explained. 4) Such seismic and faulting
activities are predicted to have reached maximum
at early postglacial time and has been decaying
since. The predicted timing and the mode of
failure of these earthquakes give excellent
agreement with the observations. Applications
1) Spatio-temporal variation of stress is useful
for locating and designing underground
repositories for radioactive waste storage,
nuclear plant safety. 2) Mitigation of
earthquake risks. 3) Mantle viscosity, which is
an important parameter in geodynamics, can be
inferred from studies of stress rotation.
48
The following is a photograph of a pressure ridge
representing the typical surface expression of
the 1989 Ungava fault rupture. The ridges are
evident because of their height, the cracked
peat, and especially the clean, light-coloured
boulders exposed. From Adams 1996 (JGR 1016193).
Used with permission.
49
Eldorado Canyon, Boulder Colorado
50
(No Transcript)
51
(No Transcript)
52
(No Transcript)
53
Location of Written Account
Epworth (Big Prairie)
54
(No Transcript)
55
(No Transcript)
56
(No Transcript)