Available online at www.sciencedirect.

com

Earth and Planetary Science Letters 269 (2008) 352 – 365

www.elsevier.com/locate/epsl

Geodetic strain across the San Andreas fault reflects elastic plate thickness
variations (rather than fault slip rate)
Jean Chéry
UNIVERSITE MONTPELLIER 2, Géosciences Montpellier, CNRS (ou CNRS/INSU), UMR 5243, Place Eugène Bataillon, CC 60 34095 Montpellier, France
Received 11 January 2007; received in revised form 15 October 2007; accepted 27 January 2008
Available online 16 February 2008
Editor: C.P. Jaupart

Abstract

The interseismic velocity field provided by geodetic methods is generally interpreted in the framework of a thick elastic lithosphere with a
slipping fault at depth. Because lateral variations of lithospheric rheology play a key role in determining the geological strain distribution, I
examine the idea that interseismic strain rate variations also occur in response to lateral variations in the elastic thickness of the lithosphere. Using
a stress balance principle and some simplifying assumptions, I show using a 1D model that elastic thickness is inversely proportional to strain rate
for the simple case of pure strike-slip faulting. Elastic thickness computed on three profiles crossing the San Andreas fault system (SAFS) suggests
that the distribution of interseismic strain rate is compatible with a thick elastic lithosphere in the Great Basin-Sierra Nevada province and on the
Pacific plate. Conversely, a thin plate with a shallow asthenosphere is needed on the SAFS to explain its high strain rate. A 2.5D Finite Element
model of interseismic strain in the Carrizo Plain region in Central California shows how known vertical and horizontal variations of elastic
properties refine 1D model predictions. In a case of a multiple fault system, I point out that the interseismic velocity is not causally tied to faults
slip rate. Therefore, analysing the velocity field across the SAFS cannot reliably provide faults slip rate distribution as previously claimed. Rather,
the apparent correlation between geologic slip rate and interseismic strain may only indicate that the elastic thickness plays a dominant role in
controlling fault strength. Finally, I suggest that interseismic geodetic strain could be a new way to infer effective elastic plate thickness on the
continents.
© 2008 Elsevier B.V. All rights reserved.

Keywords: geodesy; interseismic strain; elastic thickness; GPS; lithosphere; rheology; fault; San Andreas fault; stress

1. Introduction becoming a major tool to provide a global strain pattern at the

Earth's surface (Kreemer et al., 2003). However, switching
Seismic hazard assessment strongly relies on the measure- from interseismic strain to fault slip rate remains challenging.
ment of fault slip rates. For times longer than 10–100 kyrs, Because GPS observations are made on a small fraction of the
repeated earthquakes offset the geomorphic features crossed by seismic cycle, computing the fault slip results from a huge time
a fault (gullies and moraines). Offsets measurements and dated extrapolation using a physical model of repeated earthquakes.
features determine an average fault slip rate (Sieh and Jahns, Also, one must be confident that the surface strain are
1984). This geological approach is thought to be accurate due to representative of the bulk deformation of the continental
its direct relation to fault slip observations. However, this lithosphere, especially in the case of weak stratified plate. For
method is not always straightforward as it requires unambigu- strike-slip faulting, a widely used concept is based on a thick
ously datable offset features. Another way to compute fault slip lithosphere with an embedded fault (Savage and Burford,
rate is to measure the interseismic strain by geodetic means such 1973), referred in this paper as SB73 model or thick lithosphere
as the GPS technique. Due to the global GPS coverage, this model.
method is potentially applicable worldwide on land and is During the interseismic phase, the fault is locked from the
surface to a depth d (locking depth, see Fig. 1a). The fault below
E-mail address: [email protected]. this depth slips at a constant rate s. This model corresponds to a
0012-821X/$ - see front matter © 2008 Elsevier B.V. All rights reserved.
doi:10.1016/j.epsl.2008.01.046
 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365 353

slip faults encouraged scientists to extend the concept developed

by Savage and Burford for a single strike-slip fault to multiple
fault settings in which vertical faults delimit elastic blocks. Again,
interseismic velocities are well fit by least square inversion for
fault slip rate and locking depth. Based on its conceptual
simplicity and on successful slip rate predictions of the SB73
model, its extension to the block model is about to become a
routine tool to estimate fault slip rates in Northern California
(Freymueller et al., 1999; Savage et al., 2004; D'Alessio et al.,
2005) Southern California (Lisowski et al., 1991; Bennett et al.,
1996; Becker et al., 2005; Fay et al., 2005; Meade and Hager,
2005) and other continental areas (McClusky et al., 2000; Wallace
et al., 2004).
To summarize, the use of the thick lithosphere model is
mostly due 1) its conceptual simplicity 2) its ability to model the
geodetic strain field 3) the determination of desirable para-
meters such as the long term fault slip rate. Because the thick
lithosphere model provides a long term slip rate prediction, the
link between short term and long term time scales is mandatory
to assess this prediction reliability. Therefore, the motivation of
this paper is to examine how this model is compatible with a
long term mechanical model of lithospheric strain. Indeed, an
interseismic strain model should be viewed — in principle — as
a time fraction of a seismic cycle model. Also, a seismic cycle
model spans another time fraction embedded in a long term
evolution of the geological strain. Using the example of the
SAFS, I start with describing the mechanical and rheological
aspects of a long term model for parallel strike-slip faults. Then,
I attempt to extract the model behaviour for the different phases
of the seismic cycle (coseismic, postseismic, interseismic). This
analysis leads me to propose that the interseismic strain pattern
Fig. 1. a) The thick lithosphere model in a vertical cross-section. The vertical may be so much influenced by variations in lithosphere elastic
fault is infinite perpendicular to the cross-section. During interseismic period,
the fault above d is locked while the deep part slips at a rate s; b) adjustment of
thickness that it becomes difficult to obtain unambiguously an
the GPS velocity data (solid dots) parallel to the SAF in the Carrizo segment in estimation of the slip rate on this sole basis. I test this hypothesis
Central California (see Fig. 2 for precise location) using the screw dislocation using a mechanical analysis of the geodetic interseismic strain
model (solid curve and Eq. (1)). Values of 12 km and 34 mm/yr are used for d on the SAF using a variable thickness mechanical model.
and s respectively.
2. Geological strain, lithosphere rheology and the seismic
screw dislocation and the velocity variation at the surface is cycle on the San Andreas fault
given by:
Plate reconstruction and geological analysis have shown that
v ¼ ðs=pÞ arctan ð x=d Þ ð1Þ
most of the Quaternary strain between the Pacific plate and the
implying that interseismic velocity v reaches ~90% to the Sierra Nevada is concentrated on a few faults spaced by a few
geological fault slip rate s when the distance to the fault |x| is larger tenths of km (Brown, 1990). In zones where the SAF orientation
than ~2πd. Such a formulation is highly attractive as an entire set is aligned with the Pacific plate motion, these faults are nearly
of velocities may be fitted by adjusting the locking depth and the purely strike-slip. Therefore, the sum of the geological slip rate of
fault slip rate. Velocity profile across the central segment of the these faults is believed to be (at 10% uncertainty) equal to the
San Andreas fault (SAF) illustrates well this aspect (Fig. 1b). Sierra Nevada–Pacific plate differential motion (Dixon et al.,
Using a slip rate of 34 mm/yr and a locking depth of 12 km leads 2000). In this context, two kinds of active fault settings occur (Fig.
to a RMS misfit between the SCEC 3.0 velocity field (Shen et al., 2). First, the slip rate can be distributed over a few faults. In
2003) and the model of 2.25 mm/yr, about twice as higher as the northern California at the latitude 38–39°N, the slip occurs from
1-σ formal data uncertainty of ~1 mm/yr (Schmalzle et al., 2006). west to east on the San Andreas fault (20–25 mm/yr) the
It is remarkable that the geodetic slip rate matches well with the Maacama–Rodger Creek faults (6–10 mm/yr) and the Bartlett
geological slip rate of 34 mm/yr (Sieh and Jahns, 1984; Brown, Spring–Green Valley fault (~5 mm/yr). A similar situation occurs
1990) and that the locking depth corresponds to the maximum in southern California at a latitude of 33°N but with an inverse
seismicity depth in this zone (Miller and Furlong, 1988). The distribution with respect to the northern California setting. Indeed,
arctangent shape of the velocity profiles across many large strike- the slip rate is small on the Elsinore fault on the coast (~3 mm/yr),
354 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365

SAF (Lachenbruch and Sass, 1980; Williams et al., 2004),

suggesting high temperature in the lower crust and the upper-
most mantle induces low viscosities. Also, stress measurements
in deep boreholes indicate that the SAF supports a shear stress as
low as 10–20 MPa in the seismogenic zone (Zoback et al., 1987;
Rice, 1992) contrasting with high values of ~ 150 MPa expected
from laboratory measurements of fault friction (Byerlee, 1967).
The combination of high heat flow in a 100 km wide zone around
the SAF and the low resolved shear stress on the fault itself
provides a physical explanation for slip localization on the SAF
in a thermally weakened lithosphere (Furlong, 1993). Further-
more, the low compliance of the lithosphere seems to occur at a
smaller scale. For example, magnetotelluric experiments at
Parkfield show a low resistivity area beneath the seismogenic
part of the SAF (Unsworth et al., 1997), probably due to a high
fluid concentration related to intense shear at depth. Geomecha-
nical modelling supports this view in requiring an effective
friction coefficient of 0.05–0.17 for the Central SAF (Bird and
Kong, 1994) (Chéry et al., 2001). In zones of multiple parallel
faults such as northern and southern California, the measured
fault slip rates can be explained by a combination of lateral heat
Fig. 2. Surface velocity (blue arrows) across the SAF system given by permanent flow variations with low effective friction coefficients on faults
and campaign mode GPS data (USGS and SCEC public data) in a North American (Provost and Chéry, 2006). In general, the contribution of the
reference frame. This velocity field (black arrows) mostly represents interseismic
basal stress in mechanical modelling is neglected, although some
strain accumulation. Major faults are given in red (SAF = San Andreas fault; RC =
Rodgers Creek fault; GV = Green Valley fault; SJ = San Jacinto fault; ELS = authors have provided arguments that it may significantly affect
Elsinore fault). The three profiles across the fault system are drawn with a black the stress balance of the SAFS (Lachenbruch and Sass, 1973).
line. The surface velocity field used in the computations and in Figs. 6, 7 and 8 are Geological strain across active strike-slip faults results from a
marked with blue dots (For interpretation of the references to color in this figure combination of the rheological stratification of the lithosphere
legend, the reader is referred to the web version of this article.).
versus depth and low resisting stress in fault zones (Gilbert et al.,
1994). Depth stratification is mainly temperature dependent, and
while it is larger on the San Jacinto fault (8 mm/yr) and maximum the depth of the 350 °C isotherm represents the transition between
on the SAF to the east (22 mm/yr). In contrast, in central frictional faulting and thermally activated viscous strain (Sibson,
California between the San Francisco Bay and the Big bend, most 1982). Another transition occurs at the crust–mantle transition
of the strain occurs on the SAF only at a rate of 34 mm/yr. that marks a strength increase due to olivine rheology (Brace and
Geophysical evidence may explain this pronounced strain Kohlstedt, 1980). As both crustal and mantle viscous laws are
localization in Central California. First, a high heat flow of temperature dependent, an easy way to estimate differential stress
~ 80 mW/m2 has been measured in a zone of 100 km around the variation with depth is to assume a constant strain rate through the

Fig. 3. Typical stress envelops with depth for a) cold continental lithosphere, b) hot continental lithosphere c) weak fault zone embedded in a hot continental lithosphere
(the dashed curve represents the frictional stress associated to a high friction as in a) and b). τ represents the magnitude of deviatoric stress and :e ¼ ct means that the
strain rate does not vary with depth. Horizontal and vertical scales are indicative. The crust–mantle transition occurs at ~ 30 km depth.
 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365 355

lithosphere for different geotherms. Although the constant strain that a drastic strength reduction occurs at the brittle–ductile
rate assumption is generally incorrect in actively deforming zones transition (Gueydan et al., 2001).
as demonstrated by mechanical modelling (Chéry et al., 2001), Lithospheric strength profiles are useful to study the link
three end member models emerge for the lithospheric strength between geophysical variables such as P and T and the effective
(Fig. 3). At low heat flow (40 mW/m2), the 350 °C isotherm is rheology of the lithosphere. However, they represent the
close to Moho depth, causing the crust to be mostly brittle (Fig. maximum sustainable stress (yield stress) of the lithosphere
3a). Low temperature in the subcrustal mantle should imply very for a given strain rate, not the actual lithospheric stress. As an
high viscosities (Strehlau and Meissner, 1987). However, stress example, I consider the behaviour of the northern SAF as
controlled plastic behaviour is likely to limit the maximum modelled by Provost and Chéry (Fig. 4a,b). In their study, the
sustainable stress to about 600 MPa (Tsenn and Carter, 1987). At authors account with both strike-slip and shortening between
high heat flow (80 mW/m2), the 350 °C isotherm is shallow (7– the Pacific plate and the Sierra Nevada. The small shortening
15 km) and the Moho temperature is high (N 700 °C), leading to a strain, which is accommodated by plastic strain inside the crust
strength profile mostly controlled by upper crustal friction and and also by dip-slip fault motion, is not considered in the
middle crust viscosity (Fig. 3b). However, this model has to be present paper to keep a simple mechanical analysis. In the case
modified for a fault zone, with the constrain of low effective of pure strike-slip motion, the stress magnitude along faults
friction within the seismogenic layer (Zoback et al., 1987; Wang depends on the effective friction coefficient of the seismogenic
et al., 1995; Hassani et al., 1997) (Fig. 3c). In addition, strain rate zone and the viscosity of the middle crust. Because both
weakening and metamorphic reactions may alter the deformation frictional and viscous parameters in fault zones have low values,
processes in the middle crust, and some authors have proposed this limits the stress of the surrounding lithosphere that cannot

Fig. 4. a) Strain rate invariant associated with parallel strike-slip faults of northern California (adapted from Provost and Chéry, 2006); major faults are modelled with
intrinsically weak material and display high strain rate (Maa = Maacama fault; BS = Bartlett Springs fault); b) long term velocity in the fault-parallel direction at the
surface (solid line) and at 25 km depth (dashed line); c) fault-parallel stress profiles for the fault zone located on the SAF and for the crust. The profile to the left
corresponds to the maximum fault strength for this slip rate. The profile to the right inside the crust remains much below the maximum sustainable stress corresponding
to the dashed curve. Stress integrals with depth must be equal on the two profile to ensure stress equilibrium.
356 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365

reach its maximum value. Therefore, the lithosphere does not

deform and the viscous stress is zero due to the lack of strain. In
this case, the lithospheric stress of this zone corresponds to
elastic strain accumulation without involving frictional pro-
cesses (right profile on Fig. 4c).
Let us now consider the behaviour of such a long term model
during the seismic cycle. In contrast with the SB73 model which
is driven directly by imposing fault motion on the fault plane,
this long term model is kinematically driven from the sides of the
block at the prescribed side plate velocities. In this context,
coseismic motion occurs in response to a step in effective fault
friction, initially proposed by Brace and Byerlee (1966). Fault
rupture during large continental earthquakes takes place between
the surface and 10–20 km depth and produces elastic strain in the
fault vicinity. Postseismic motion occurs minutes to years
following the earthquake as a result of a variety of processes like
afterslip of the deep fault plane, viscoelastic processes in the
middle crust and the mantle and poroelastic effects in the crust. If
I ignore the poroelastic effect for this discussion, both afterslip
and viscoelastic strain act to release stress buildup at depth and
reload the fault plane above. Due to the thermal stratification of
the lithosphere, a large viscosity spectrum [1017–1019 Pa s] is Fig. 5. Interseismic stress accumulation and corresponding velocity. a) Cross-
likely to control the strain following large earthquakes as shown section of a lithosphere with lateral variation along x-axis of its geodetic elastic
by postseismic modelling studies (Freed et al., 2006). Once this thickness Tg. The position of the 350 °C isotherm is given by the dashed line and
marks the upper limit of the viscoelastic zone. A low deviatoric stress is assumed
stress transfer is complete, a steady interseismic strain build up to occur below corresponding to effective viscosities lower than 1019–1020 Pa s.
occurs in response to plate motion. A chief difference of this A fault-parallel stress increase Δτxy occurring a time interval Δt (see Eq. (2)) is
phase with respect to the long term behaviour is that the fault is shown on two profiles. For each profile, the stress increase is assumed to be
locked, implying that the whole seismogenic layer behaves constant with depth accordingly and is materialized with a solid rectangle. Due
elastically or viscoelastically with viscosities higher than 1021 Pa to stress equilibrium the stress increase integrals on the two profiles are equal. b)
Corresponding interseismic velocity field (y-axis component) across the
s. At depth, a low viscosity stress occurs in response to lithosphere computed using Eq. (3).
interseismic strain beneath the seismogenic zone.
If one attempts to interpret the interseismic strain across the
SAFS with the model of Fig. 4c, it becomes clear that the the mantle by postseismic modelling are on the order of 1019 Pa
surface strain has to be influenced by the thickness of the elastic s, suggesting that these zones produce a stress contribution of
and viscoelastic layers. As suggested by Fig. 5, thin parts of the ~ 1 MPa if an interseismic strain rate of 10− 14 s− 1 is applied.
layer should display strain accumulation while thick parts To summarize this analysis, interseismic strain measured by
should not accumulate much strain. Although this two-layer geodetic tools at the Earth's surface may reflect the elastic strain
model has been invoked for decades to explain postseismic accumulation of the seismogenic zone (i.e., grossly the part of
strain (Nur and Mavko, 1974; Pollitz et al., 2000) (Kenner and the crust above 350 °C and possibly the cold uppermost mantle)
Segall, 1999), it is less used to explain the interseismic strain and the viscoelastic behaviour of a thin zone below the
distribution across fault systems (Bourne et al., 1998; Cohen seismogenic layer in the crust and possibly in the mantle. To
and Darby, 2003; Schmalzle et al., 2006). In such a case, elastic significantly contribute to interseismic stress accumulation, the
strength is likely to depend on the product of the thickness of the corresponding viscosities must be higher than 1021 Pa s. This
seismogenic zone and the average shear modulus of this layer as way to interpret interseismic strain has been proposed for the
detailed later. The viscoelastic strength determination follows SAFS and for the Alpine fault of New Zealand (Bourne et al.,
the same scheme and depends on the product of the thickness of 1998) but assuming that viscous strain in the lower crust and in
the viscous layer and its average viscosity. Two lines of the mantle drives and controls the entire fault system. Rather, I
evidence suggest that the viscous contribution to the litho- propose that interseismic strain reflects lateral variations in the
spheric strength is relatively small. First, a large stress reduction thickness and elastic modulus of a lithospheric stress guide. This
with depth occurs in only a few km only due to the high view is obviously close to the elastic plate model used to explain
sensitivity of power law flow to temperature. For example, a the flexural behaviour of the lithosphere (Watts, 2001). The main
typical granite-type power law (Kirby, 1985) loaded at a strain difference comes from the kind of forces applied to the plate.
rate of 10− 14 s− 1 implies a deviatoric stress of 100 MPa at Classical plate theory aims to explain the relationship between
350 °C and only 1–10 MPa at 450 °C. Assuming a surface vertical motion (plate bending) and plate thickness, while the
thermal gradient of 20 °C/km, this suggests that the thickness of concept I propose here considers the relation between horizontal
the layer hosting significant viscous stress should not exceed 5– strain and plate thickness. Because this thickness is evaluated
10 km. Second, effective viscosities determined in the crust and by analysing horizontal geodetic strain, I name it geodetic elastic
 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365 357

thickness (GET) or Tg in order to differentiate it to the flexural space derivative of the velocity field, I compute the strain rate
elastic thickness (Te) provided by plate bending analysis. using a least square adjustment of the interseismic velocity over
a moving window of 20 km minimum half-width. Therefore,
3. A simple model between strain rate and elastic thickness most of the long and short wavelength features of the profiles are
preserved (Fig. 6a) and interseismic strain rate is given by the
To build a simple model explaining the relation between the slope of the least square adjustment (Fig. 6b). The given formal
interseismic strain and the lithospheric strength, I assume here error on the strain rate corresponds to the slope uncertainty
that the crust and the mantle are elastic for temperature lower associated to the least square adjustment. This leads to obtain a
than respectively 350 °C and 750 °C. Also, I neglect small error when all the points are aligned in the moving
viscoelastic effects that is to say that the stress at temperature window, therefore corresponding to a constant strain rate model.
greater than 350 °C and 750 °C for the crust and the mantle A more complete formulation that should also include individual
respectively is negligible. The contribution of interseismic RMS associated to the data has not been developed here. The
strain corresponds therefore to the deformation of a plate having computation of the elastic thickness using Eq. (3) requires a
an effective thickness Tg. as presented in Fig. 5a. Because the priori information on Tg because a trivial solution is given by
mechanical system obeys stress equilibrium and assuming that Tg = 0, corresponding to the lack of information on the shear
only the stress component τxy contributes to the total force, force change of Eq. (2). Because earthquake seismicity near the
stress increase Δτxy during a time Δt on two profiles of SAF does not occur below 12–15 km, I assume a minimum
thickness Tg1 and Tg2 satisfies the relation:
Z Tg1 Z Tg2
Ds1xy ð zÞdz ¼ Ds2xy ð zÞdz ¼ DF ð2Þ
0 0

where ΔF represents the shear force change applied to the

lithosphere. I assume a linear shear stress–shear strain relation
given by Δτxy = G·Δεxy where G is the average shear modulus
on the layer. Assuming that Δεxy is constant along a vertical
profile and dividing the previous equation by Δt leads to:
P :
G ð xÞ  Tg ð xÞ  e xy ð xÞ ¼ C ð3Þ
P
where G is the average shear modulus on the profile and C is a
constant, meaning that fault-parallel strain rate e: xy is inversely
P
proportional to the integrated elastic strength G  Tg . With the
simplifying assumptions above, knowledge of the interseismic
strain rate should teach us how elastic thickness may vary. I test
this hypothesis on three profiles across the SAFS.

4. Analysis of velocity profiles crossing the San

Andreas fault

GPS profiles shown on Fig. 2 cross the SAFS north of San

Francisco Bay (North Bay profile, see Fig. 6), in the Carrizo
plain south of Parkfield (Carrizo profile, see Fig. 7) and close to
the Salton Sea south of Los Angeles (Salton profile, see Fig. 8).
Faults are straight around these three locations so the deforma-
tion is nearly two-dimensional. Also, no large earthquakes have
affected the SAF since the 1906 San Francisco earthquake for
the North Bay profile and the 1857 Ft. Tejon earthquake for the
Carrizo profile. Also, there is no evidence that the 1857 rupture
extended down into the Salton Trough region. Therefore, the
Fig. 6. Interpretation of the GPS velocity field on a profile perpendicular to the
postseismic signal on the three profiles is expected to be small northern SAFS. a) Fault-parallel GPS interseismic velocity fields (see Fig. 2 for
compared to the interseismic strain. I use the public data location). A continuous version of the velocity measurements is given by the red
available on USGS and SCEC web sites corresponding to GPS curve. The RMS (mm/yr) of the adjustment between the curve and the data is given
surface velocities with respect to the North American plate. by the blue solid line with the scale to the right; b) fault-parallel horizontal strain
Because the RMS associated to the data are of the order rate corresponding to the continuous velocity field slope c) elastic thickness across
the SAFS based on Eq. (3) (black curve) and on the assumptions given in the text.
1 mm/yr, the direct derivative of the linear velocity field Minimum and maximum thickness associated to the strain rate formal error are
interpolation with respect to the profile direction is meaningless represented by red lines (For interpretation of the references to color in this figure
and not of a practical use in our case. In order to obtain a smooth legend, the reader is referred to the web version of this article.).
358 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365

elastic thickness on each profile of 13 km. Elastic thickness

distribution is computed according to this value, also assuming
that the average shear modulus does not vary along the profile. If
computations were done with a different minimum elastic
thickness, it would have been affected elastic thickness on the
profile only by a multiplicative constant (see Eq. (3)). From
north to south, interseismic strain rate patterns and their
corresponding elastic thickness are markedly different.
Along the North Bay profile, strain rate displays a marked
asymmetry with a maximum along the Pacific plate to the west
and a slow decrease towards the east up to the Central Valley and
the Sierra Nevada. Consequently, the elastic thickness is higher
close to the Pacific plate (40–70 km) and gently decreases (13 to
20 km) across the Coast Range from the SAF to the Green Valley
fault. Because of the small strain rate of the Central Valley, the
inverse relation between strain rate and thickness leads to an
elastic thickness larger than 100 km (Fig. 6c).
Interseismic velocity in central California leads to a different
interseismic strain variation (Fig. 7b). Strain rate gently increases
from southwest (the Pacific coast) to northeast up to a maximum
10 km east to the SAF trace as noticed by previous work
(Schmalzle et al., 2006). Despite a limited GPS data set in the
Central Valley and Sierra Nevada, the strain rate 20 km east to the
SAF is virtually zero according to other geodetic studies of the
Sierra Nevada block (Dixon et al., 2000). According to this strain

Fig. 8. Same as for Fig. 6 for the southern SAFS (see Fig. 2 for location).

variation, elastic thickness gradually decreases from values higher

than 80 km to 13 km across the western Coast Ranges (Fig. 7c), and
jumps back to large values east to the SAF in the Central Valley.
The southern California profile near the Salton Sea reveals a
reversed strain pattern compared to the one obtained in the
North Bay area. Elastic thickness progressively decreases from
the Pacific coast when crossing the Elsinore fault, reaches its
minimum value between the San Jacinto and SAF and increases
towards ~80 km on the north American plate to the east (Fig. 8c).

5. Finite Element model of the interseismic strain

The simple formulation used to compute the elastic thickness

of the SAFS provides a first order relation between interseismic
strain and plate thickness. However, stress components are
likely to vary vertically within the plate, implying that an
accurate strain solution requires solving a more complete stress
balance equation. Also, the assumption of constant elastic
modulus through the entire lithosphere is questionable, as
vertical and lateral elastic properties variations of the litho-
sphere are known to be significant (Meissner, 1986). Involving
these effects with solving stress balance equation with complex
geometry needs to be done numerically. I design one experiment
to determine if the simple concept developed above about the
strain rate–elastic thickness relation still holds for a finite
Fig. 7. Same as for Fig. 6 for the central SAFS (see Fig. 2 for location). thickness lithosphere. Because errors induced by the 1D model
 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365 359

:
Fig. 9. a) Horizontal and vertical Young's modulus variation used in the Finite Element model; b) Geometry of the FEM and stress rate accumulation s xy corresponding
to a 34 mm/yr loading velocity between the lateral sides (at 0 and 350 km). Zone of low stress rate to the right corresponds to the Great Valley and the Sierra Nevada.
Zone 20 km east to the SAF has low elastic modulus based on seismological evidence and loads at a high stress rate of 0.01 MPa/yr. Note that this value predicts an
interseismic stress loading of 2.5 MPa for a recurrence time of 250 yr, which is compatible with the average static stress drop for a large earthquake (Hanks, 1977).
Zone west to the SAF corresponds to a progressive plate thickening increase.

are likely to be higher when large geometrical variations occur, I smoothing of the discrete velocity data, therefore representing
use the example of the central SAF for which high thickness the best possible data fit. Despite its lower accuracy, I argue that
gradients are expected. the data adjustment provided by the FEM has a greater physical
The geometry of the model represents a cross-section of the meaning than the 1D model presented before. This opinion is
elastic part of lithosphere perpendicular to the fault direction. The based on three factors. First, it corresponds to a finite thickness
loading corresponds to a motion of the Pacific plate with respect lithosphere, implying that the hypothesis of having :s xy constant
to a fixed Sierra Nevada at a rate of 34 mm/yr parallel to the fault
direction. Assuming no velocity variation in the direction parallel
to the fault direction (y), non-zero stress components are τxy and
τyz. Both horizontal and vertical bulk elastic variations are taken
into account (Fig. 9a). I incorporate Young's modulus increase
with depth as observed worldwide in continents (Meissner, 1986).
Based on a seismic velocity model of the Parkfield area (Eberhart-
Philips and Michael, 1993) I also account for a lateral decrease of
Young's modulus in a 20 km width zone east of the SAF. I also
adjust elastic thickness along the Carrizo profile. Elastic thickness
of the Great Valley–Sierra Nevada is set to 200 km in order to
account for both its low geodetic deformation (Dixon et al., 2000)
and its high flexural rigidity (Kennelly and Chase, 1989). Elastic
thickness of the Pacific plate west to the SAF is set to 40 km to
make it compatible with currently estimated elastic thickness
(Watts, 2001).
The interseismic profile velocity for the model of Fig. 9b
displays a data-model RMS of 1.77 mm/yr (Fig. 10). Due to
intrinsic velocity errors of about 1 mm/yr on the Carrizo profile,
the smooth interseismic curve (1D model of Fig. 6) fits velocity
Fig. 10. Interseismic velocity field provided by the Finite Element model (thick
data with a RMS of 1.34 mm/yr. The data fit of the FEM model is black line) compared with the smooth data fit (dashed thin line) and with the
not as good as the one provided by the 1D model. This is thick lithosphere model (purple curve). Discrete GPS velocity values across the
understandable as the 1D strain model is obtained by direct central SAFS are given by the solid black circles.
360 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365

on a vertical profile is no longer needed. Clearly, this 1D 1977). Indeed, large lateral variations of bulk rigidity are likely to
assumption is violated when high thickness gradients are present occur and could explain the asymmetry of the deformation.
as shown by Fig. 9b. Second, the FEM allows us to include our However, a bulk rigidity contrast of 10 that is required to explain
best a priori knowledge of the elastic properties of the the strain asymmetry around the Sumatra fault (Le Pichon et al.,
lithosphere based on seismology. Interestingly, a reliable 2005) probably exceeds known variations of shear modulus
estimate of bulk elastic properties requires the thickness only within the crust. An alternative explanation for strain asymmetry
to be adjusted. Third, interseismic strain provided by the FEM is a variation of the elastic thickness in conjunction with a bulk
results from a trial-and-error procedure. It is likely that a suitable rigidity contrast (Melbourne and Helmberger, 2001; Cohen and
inversion technique based on a grid search for an optimal elastic Darby, 2003; Schmalzle et al., 2006).
thickness would lead to a fit as good as the one provided by the
1D smooth fit. In comparison to the 1D and the FEM data fit, the 6.1. Interseismic strain, elastic plate thickness and fault slip rate
adjustment of fault slip rate and locking depth of the SB73 model
leads to a data fit of 2.25 mm/yr. This is mostly because the As shown by the analysis of the central SAF profiles, SB73
asymmetry of the elastic profile resulting from the variable and variable thickness models both provide a good fit to the
elastic thickness model fits the GPS data better compared to the interseismic velocity field. In contrast, different behaviours can
symmetric solution provided by the SB73 model. However, the be expected between these models when fault slip rates are
fit of the thick lithosphere model is quite acceptable given the searched. In the case of the SB73 model, the long term fault slip
small number of free parameters and may be still improved if rate is equal to the far-field velocity (the differential plate
elastic modulus contrast across the SAF is taken into account. velocity). The long term slip rate for the variable thickness
model is less straightforward to define because of the remote
6. Discussion fault drive. As discussed in the introduction, the effective fault
friction should be considered here. If the fault friction is low, the
Choosing among different mechanical models of interseismic fault will slip at the differential plate velocity because the
strain on the sole basis of geodetic data fitting has been shown to lithosphere does not deform anelastically. If the fault friction is
be meaningless because of the non-uniqueness of the problem high, two cases have to be considered. If one assumes that the
even in the case of a single fault (Savage, 1990). In other words, a lithosphere always behaves elastically (meaning that it can
bad data fit can allow us to discard a model but a good data fit is sustain arbitrarily high strain) the fault slip rate must still be
not proof of a model's relevance. Therefore, in chasing among equal to the differential plate velocity. However, both in-situ and
possible models I need to consider the problem from a broader laboratory rock strength estimates indicate that the differential
point of view than the one of finding the best data. Rather, it stress of the seismogenic lithosphere is limited by a friction
requires consideration of each model's relevance from a coefficient of [0.6–0.8] (Townend and Zoback, 2000). There-
geophysical, rheological and geodynamical point of view. Having fore, a high fault friction is likely to pervasively deform the
in mind the current view of the thermomechanic state of the lithosphere around, especially where the plate is the thinnest. For
lithosphere beneath the SAF (Lachenbruch and Sass, 1980; this case only, the fault slip rate is expected to be different to the
Furlong, 1993; Chéry et al., 2001), the use of a thick elastic model differential plate velocity. Considering that large faults separat-
for the SAFS seems at odds to this knowledge. For example, the ing thick lithospheric plates are often thought to sustain low
northern SAFS is very juvenile as it results from the northward shear stress, the fault slip rate determined by the variable
progression of the Mendocino triple junction (Furlong, 1984). thickness model must therefore be in fair agreement with the best
There, the disappearance of the subducting plate is thought to have fit SB73 model.
opened an asthenospheric window in contact with the upper plate.
As a consequence, a sharp contrast is likely to exists between a 6.2. Slip rate computation of a parallel fault system
thin lithosphere (10–20 km) surrounding the SAFS and a thick
lithosphere (50–100 km) to the east in the Great Valley and in the From a mechanical viewpoint (see Appendix for a discus-
Sierra Nevada. In other words, an interpretation of the whole fault sion), the SB73 model mimics the mechanical behaviour of two
system using a thick elastic lithosphere is therefore unlikely thick lithospheric blocks separated by a low strength fault zone.
because of the presence of a hot low viscosity zone in the upper Interestingly, most fault slip rates inferred from SB73 and block
mantle in the region of the slab window. By contrast, the variable models analysis around continental plate boundaries are in good
elastic thickness model predict a low rigidity of the SAFS, which agreement with the corresponding geologic slip rates (Reilinger
is compatible with a shallow low viscosity zone. As already noted et al., 2006). Therefore, it is logical to conclude that the geodetic
(Le Pichon et al., 2005; Schmalzle et al., 2006), the data fit fault slip determination using the SB73 model is valid only if it
provided by the Savage and Burford model is limited by the corresponds to a setting of a weak fault between two strong
asymmetrical character of the velocity field. In the simple case of a plates. However, many other fault settings occur in nature as
single vertical strike-slip fault, one expects a perfect symmetry multiple fault system, faults embedded in wide orogens or high
with respect to the fault axis. This is not always the case, as shown plateaus. To discuss this point in the framework of the SAFS, I
by Le Pichon et al. for different strike-slip faults. A plausible attempt here to explain how the variable thickness model
explanation is that lateral variations of elastic properties of a thick behaves when two faults or more are embedded (Fig. 11). This
lithosphere cause a symmetry break (Rybicki and Kasahara, model is directly adapted from the 1-fault model of Fig. A1c but
 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365 361

Fig. 11. Mechanical model of a variable thickness lithosphere model with two embedded fault. Each fault strength i is the sum of a frictional stress occurring on a width
hif and of a viscous stress occurring on a width hiv.

adding another frictional discontinuity and a viscoelastic zone This relation is represented on Fig. 12 for different values of
according to Fig. 4c. h1 s1 in the case or h2 N h1. As the respective contribution of
As shown by the previous mechanical modeling, frictional frictional and viscous fault strength is still a matter of debate,
and viscous fault strength both control slip rate distribution in a this formulation is well suited as it does not require an a-priori
multiple fault system (Bird and Baumgardner, 1984; Provost choice. In order to see how southern and northern SAFS behave
and Chéry, 2006). Using the example of Fig. 11 and using a 1D with respect to Eq. (8), let us assume that viscous to frictional
balance assumption similar of Eq. (2), it can easily be shown ratio is the same for each fault. Thus the ratio between frictional
that the effective strength of both faults are equal. Also, the thicknesses is equal to the ratio given by geodetic elastic
strength of the fault i is equal to the sum of the frictional thicknesses. Slip rate and frictional thickness ratios are given in
strength Ffi and of the viscous strength Fvi. For each fault, the Table 1 and plotted in Fig. 12. Using preferred values for slip
frictional strength is equal to the stress integral along the rates, only pairs including the Elsinore fault display θ1s1 values
frictional thickness hfi. Assuming a frictional shear stress larger than 3. For other pairs, elastic thickness and slip rate
constant with depth leads the frictional strength to be equal to: ratios match curves computed with between 1/3 and 3,
suggesting that viscous and frictional strengths may have an
Ffi ¼ si hif ð4Þ equal influence on the SAFS. In order to see how this parameter
compares with an a priori calculation, I use the following set of
The viscous strength is equal to the integral of the viscous parameter: a viscosity of 1019 Pa s; a fault zone width of 1 km; a
stress along the viscous layer which has a thickness hvi. This ratio hivhif = 0.25; a frictional shear stress of 10 MPa. In such a
integral depends on the average strain rate inside this layer
around the fault plane and is also affected by the large viscosity
variations that likely occur with depth. Assuming that the
viscous strain occurs on an average width w, the viscous
strength is therefore defined by the relation

Fvi ¼ giw hiv  si ð5Þ

where is the fault viscosity η divided by w and s the fault slip
i i

rate. The total fault strength is equal to

 
F i ¼ si 1 þ hi si hif ð6Þ

with

giw him
hi ¼ ð7Þ
si hif

Therefore, θi represents the viscous component of the fault

strength. Assuming for purposes of discussion that both θi and τi
are equal for both faults, the slip rate ratio is given by:
"   # Fig. 12. Normalized relation between slip rate and mechanical thickness using
s2 1 h1f 1 Eq. (8) for different values of the viscosity parameter (red lines). Black circles
¼ max 0; 1 þ 1  ð8Þ
s1 h s1 h2f h1 s1 give the relation between slip rate and thickness as computed in Table 1. Black
lines represents slip rate uncertainties.
362 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365

Table 1 knowledge of the SAFS suggests that a thin elastic lithosphere

Geodetic elastic thickness (provided by this study) and known fault slip rates for indicates a locally weakened plate. This kind of weakness can be
northern (lines 1 to 3) and southern (lines 4 to 6) SAFS (Shen-Tu, Holt et al., 1999)
thermally induced, as high temperature gradients reduce the
Fault1–fault2 T1g T2g s1 s2 h1f /h2f s2/s1 thickness of both brittle and viscous layers. This is likely to occur
(km) (km) (mm/yr) (mm/yr)
in mountain belts, rifts or extending plateaus as shown by
SAF–RC (a) 13 18 20 [12–30] 9 [7–10] 0.72 0.45 [0.30–0.75] geophysical measurements and mechanical modelling (Gaude-
SAF–GV (b) 13 20 20 [12–30] 6 [4–8] 0.65 0.30 [0.20–0.50]
mer et al., 1988; Buck 1991; Cattin et al., 2001). Also, a large
RC–GV (c) 18 20 9 [7–10] 6 [4–8] 0.90 0.66 [0.44–0.88]
SAF–SJ (d) 13 17 29 [10–35] 12 [8–24] 0.76 0.41 [0.27–1.00] contribution to the lithospheric weakness could be related to low
SAF–ELS (e) 13 33 29 [10–35] 6 [2–9] 0.39 0.21 [0.06–0.6] strength faults that decouple (in a stress meaning) adjacent
SJ–ELS (f) 17 33 12 [8–24] 6 [2–9] 0.51 0.50 [0.16–0.75] plates. Subduction faults and large intracontinental faults are
Numbers in columns 5, 6, 8 represent preferred values while brackets indicate probably a good example of this kind of weakness (Wang et al.,
min/max estimates. SAF = San Andreas fault; RC = Rodgers Creek fault; GV = 1995; Hassani et al., 1997; Cattin et al., 2001).
Green Valley fault; SJ = San Jacinto fault; ELS = Elsinore fault. Up to now, elastic plate thickness has been computed for
continental and oceanic plates using the idea that vertical loads
case, θi = 2.5 108 s. Using a velocity of 25 mm/yr for the SAF applied to the lithosphere produce vertical motions by plate
(7.92 10− 10 ms− 1), is close to 0.2. bending (Watts, 2001). Using stress equilibrium of the plate and
Given the large uncertainty in rheological parameters such as isostatic assumption, knowledge of the horizontal distribution of
the fault zone viscosity and width, the agreement between this the loading function (topography, internal loads, mantle buoy-
a-priori computation and the value deduced from the elastic ancy, glaciations, etc) permits determination of the flexural
thickness and slip rate plot is encouraging. rigidity that best explains topographic and gravimetric signals.
This analysis suggests that several geometrical and rheological Knowledge of the elastic parameters of the lithosphere permits
parameters control fault strength and therefore geological fault conversion of flexural rigidity to equivalent plate thickness Te.
slip rates. Frictional strength is controlled by the fault friction and However, one needs to be careful when plate thickness are
the seismogenic thickness, while viscous strength is controlled by obtained with different kinds of loads. Because the model
the viscosity and the size of the viscous domain. Any variation of assumes an elastic plate over an inviscid fluid, the determination
these rheological parameters on a fault will cause the slip rate to of the flexural rigidity in nature is correctly done if the load is
vary as already demonstrated by some studies (Roy and Royden, applied for a time scale long enough to allow a complete stress
2000; Provost and Chéry, 2006). The present analysis suggests a relaxation at depth. Short time scale loading does not allow such
complex link between geologic slip rate and interseismic strain if stress relaxation, therefore leading to larger values of equivalent
more than one fault are active. Despite that the SB73 model is plate thickness. For example, mechanical modelling of the rapid
based on a direct and causal relation between interseismic strain filling of the lake Mead in the Basin and Range leads to a plate
and fault slip rate, these two parameters may only be positively thickness estimate of 20–30 km (Kaufmann and Amelung,
correlated in nature. In order to properly interpret their relation 1995) much higher than the 5–15 km thickness found using the
one must understand their causal relation. Indeed, fault slip rate surface topography as a loading function (Lowrie and Smith,
chiefly depends on effective fault friction, fault zone viscosity and 1995). On both oceanic and continental plates, a clear correlation
the mechanical thickness of all these process zones. Because a is found between elastic thickness and the thermal state of the
greater thickness induces a strength increase if other rheological lithosphere if a long term loading is considered. The 500–550 °C
parameters remains unchanged, such an increase in thickness isotherm matches the lower limit of the elastic plate in oceans,
causes both the slip rate and the interseismic strain to decrease. while no strong correlation has been found between effective
But a change of effective fault friction modifies fault slip rate elastic thickness and some particular isotherm for continental
without changing the interseismic strain. Therefore, the apparent crust (Watts, 2001). The current interpretation for oceanic
correlation between geologic slip rate and interseismic strain may lithosphere is that most of the elastic stress is stored between the
only indicate that the elastic thickness plays a dominant role in surface and the 500–550 °C isotherm. In the case where elastic
controlling fault strength. stresses reach the frictional strength of the lithosphere (generally
due to high plate curvature as it occurs in subduction zones), an
6.3. A global relation between elastic thickness and interseismic elastoplastic plate bending model has to be used (Judge and Mc
strain? Nutt, 1991) to account for a reduced elastic thickness. The
situation is more complex in a moderately cold continental
The goal of this paper is to study the interseismic strain rate lithosphere for which a large part of the strength is probably
distribution across the SAFS and its relation to lateral rigidity stored in the uppermost mantle. As I show using the example of
variations. If such a relation holds for this plate boundary, the SAF, the use of horizontal GPS velocity gradients at the
perhaps a similar relation can be expected for other deformation Earth's surface through a stress equilibrium principle is another
areas such as subduction zones, mountain belts, continental or way to estimate effective plate thickness using its transverse
oceanic rifts. If a relation between interseismic strain and elastic rigidity as a parameter to invert. However, such a use of GPS
thickness is assumed, then plate boundaries that displays high velocities relies on two key assumptions. First, the strain must
strain rate concentration would need to be interpreted like thin really represent the interseismic stage during which faults are
elastic zones. Is this mechanically understandable? Current locked. The second assumption is that the plate strain is the result
 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365 363

of stress equilibrium inside the plate without coupling with the Appendix A
mantle other than hydrostatic forces. For example, horizontal
forces applied at the base of the elastic plate such as deep The variable elastic thickness model has markedly different
convection or slab traction can induce strain on the surface, seismic hazard implications compared to those deduced from
therefore leading to an incorrect plate thickness estimate. SB73 and block models. For the latter, fault slip rates are a free
Keeping in mind that the relation between interseismic strain parameter to adjust. For the former, elastic thickness can be
and plate thickness may therefore breakdown, the applicability computed, but no direct link can be made with the long term fault
of this theory on continents is worldwide. Indeed, the rapid slip rate. This lack of information is clearly related to the model's
growth of GPS and InSAR mapping is already providing a dense assumption: interseismic strain is interpreted on the time scale
and accurate velocity fields in many active region. Suitable during which geodetic measurements are made. By contrast with
interpolation of this velocity field makes possible the strain rate the SB73 model, no hypothesis is made about the relation
computation (Kreemer et al., 2003). Using an inversion between short term interseismic strain and long term slip rate.
procedure, this would potentially allow computation of rigidity Does this mean that interseismic strain cannot be used to predict
maps on continents. fault slip rate? To discuss this important issue, let us consider
again the formal differences between these two models. The
7. Conclusion most significant is the way to apply the boundary conditions to
the model. In the case of SB73 model, the differential velocity is
The idea that interseismic strain is linked to variable elastic applied on the deep part of the fault (near field boundary
plate thickness is markedly different than the usual geodetic condition, see Fig. A1a). In such a case, the fault slip rate is a
data interpretation using the block model driven by fault slip at kinematical parameter and has no relation to the fault strength. In
depth. I have argued that the former model is rheologically more the case of the variable elastic thickness model, the velocity
plausible as it is directly linked to brittle and ductile lithosphere boundary condition is remotely applied according to differential
properties. In such a model, faults do not play a direct role plate motion. Because the source of plate motion is thought to be
during the interseismic phase because they are locked. controlled by large-scale body forces, this way to setup the
Interseismic strain is therefore chiefly controlled by elastic boundary condition is probably wiser than a local drive. Let us
plate properties (mostly bulk elastic modulus and plate
thickness). Two main implications can be drawn:

– the use of the block model to infer fault slip rate must be
restricted to single faults embedded in high rigidity domains
like for example the central north Anatolian fault or the
central SAF. Incorrect slip rates are likely to be inferred if
this condition is not fulfilled. This case occurs when multiple
faults are present as in southern and northern California and
suggests that such geodetically based fault slip rates could be
unreliable.
– In a simple tectonic setting such as pure strike-slip faulting,
high interseismic strain can be interpreted in the same way as a
low rigidity (small elastic thickness) zone, while low strain
would correspond to a higher rigidity (large thickness). If this
1D analysis can be extended to 2D based on stress equilibrium
principle, the variable plate rigidity model could be applied to
compute elastic thickness on continents with a suitable
inversion procedure using interseismic strain as input data.

Acknowledgements

This paper benefited from several discussions with my

colleagues. I particularly thank Xavier Le Pichon who forced
me to clarify my interpretations. Philippe Vernant read an early
version of the paper. Careful reviews of Wayne Thatcher and of
two anonymous reviewers allowed to significantly improve the
manuscript. I thank also Jessica Murray for helping me in
gathering USGS geodetic data. SCEC and USGS are acknowl-
Fig. A1. Differences and similarities between the thick lithosphere model and
edged for providing geodetic data of high quality. Generic the variable thickness lithosphere model; a) thick elastic lithosphere model
Mapping Tool (GMT) software has been used to prepare most (SB73 model); b) modified SB73 model with remote boundary conditions c)
of the figures. variable thickness elastic lithosphere.
364 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365

now imagine how to reproduce the interseismic velocity field of Fay, N.P., Humpreys, E.G., et al., 2005. Fault slip rates, effects of elastic
the SB73 model using a remote boundary condition (Fig. A1b). heterogeneities on geodetic data, and the strength of the lower crust in
the Salton Trough region, southern California. J. Geophys. Res. 110.
Fig. A1 Differences and similarities between the thick doi:10.1029/2004JB003548.
lithosphere model and the variable thickness lithosphere model; Freed, A.F., Burgmann, R., et al., 2006. Stress-dependent power-law flow in the
a) thick elastic lithosphere model (SB73 model); b) modified upper mantle following the 2002 Denali, Alaska, earthquake. Earth Planet.
SB73 model with remote boundary conditions c) variable Sci. Lett. 252, 481–489.
Freymueller, J.T., Murray, M.H., et al., 1999. Kinematics of the Pacific–North
thickness elastic lithosphere.
America plate boundary zone, northern California. J. Geophys. Res. 104
If the fault strength is high, the slip will be not allowed and the (B4), 7419–7441.
strain would be homogeneous throughout the model. Conversely, Furlong, K.P., 1984. Lithospheric behavior with triple junction migration: an
the way to simulate the velocity jump on the deep fault is to assign example based on the Mendocino triple junction. Phys. Earth Planet. Inter.
a low strength for this part of the fault. Therefore, the low strength 36, 213–223.
of the deep part of the fault allows to interpret the SB73 model as Furlong, K.P., 1993. Thermal–rheologic evolution of the upper mantle and
the development of the San Andreas fault system. Tectonophysics 223, 149–164.
the juxtaposition of two thick blocks only coupled by the locked Gaudemer, Y., Jaupart, C., et al., 1988. Thermal control on post orogenic
zone above the velocity jump. This mechanical view is equivalent extension in collision belts. Earth Planet. Sci. Lett. 89, 48–62.
to a variable thickness model in which two thick blocks are Gilbert, L.E., Scholz, C.H., et al., 1994. Strain localization along the San
separated by an asthenospheric channel (Fig. A1c), and these Andreas fault: consequences for loading mechanics. J. Geophys. Res. 99
three models should display similar interseismic surface velocity (B12), 23,975–23,984.
Gueydan, F., Leroy, Y., et al., 2001. Grain-size-sensitive flow and shear stress
field (although they are not identical). enhancement at the brittle–ductile transition of the continental crust. Int. J.
Earth Sci. 90, 181–196.
References Hanks, T., 1977. Earthquake stress drop, ambient tectonic stress, and stress that
drive plate motions. PAGEOPH 115, 441–458.
Becker, T.D., Hardebeck, J.H., et al., 2005. Constraints on fault slip rates of the Hassani, R., Jongmans, D., et al., 1997. Study of plate deformation and stress in
southern California plate boundary from GPS velocity and stress inversion. subduction processes using two-dimensional numerical models. J. Geophys.
Geophys. J. Int. 160, 634–650. Res. 102, 17951–17965.
Bennett, R.A., Rodi, W., et al., 1996. Global positioning system constraints on Judge, A.V., Mc Nutt, M.K., 1991. The relationship between plate curvature and
fault slip rates in southern California and northern Baja, Mexico. J. Geophys. elastic plate thickness: a study of the Peru–Chile trench. J. Geophys. Res.
Res. 101, 21943–21960. 96, 16625–16640.
Bird, P., Baumgardner, J., 1984. Fault friction, regional stress, and crust mantle Kaufmann, G., Amelung, F., 1995. Reservoir-induced deformation and
coupling in Southern California from finite elements models. J. Geophys. continental rheology in vicinity of Lake Mead, Nevada. J. Geophys. Res.
Res. 84, 1932–1944. 105, 16341–16358.
Bird, P., Kong, X., 1994. Computer simulations of California tectonics confirm Kennelly, P.J., Chase, C.G., 1989. Flexure and isostatic residual gravity of the
very low strength of major faults. Geol. Soc. Amer. Bull. 106, 159–174. Sierra Nevada. J. Geophys. Res. 94, 1759–1764.
Bourne, S.J., England, P.C., et al., 1998. The motion of crustal blocks driven by Kenner, S., Segall, P., 1999. Time-dependence of the stress shadowing effect and
flow of the lower lithosphere: implications for slip rates of faults in the south its relation to the structure of the lower crust. Geology 27 (2), 119–122.
island of New Zealand and Southern California. Nature 391, 655–659. Kirby, S.H., 1985. Rocks mechanics observations pertinent to the rheology of the
Brace, W.F., Byerlee, J.D., 1966. Stick-slip as a mechanism for earthquakes. lithosphere and the localization of strain along shear zones. Tectonophysics
Science 153, 990–992. 119, 1–27.
Brace, W.F., Kohlstedt, D.L., 1980. Limits on lithospheric stress imposed by Kreemer, C., Holt, W., et al., 2003. An integrated global model of present-day
laboratory experiments. J. Geophys. Res. 85, 6248–6252. plate motions and plate boundary deformation. Geophys. J. Int. 154, 8–34.
Brown, R.D.J., 1990. Quaternary deformation. In: Wallace, R.E. (Ed.), The San Lachenbruch, A.H., Sass, J.H., 1973. Thermo-mechanical aspects of the San
Andreas Fault System, California, vol. 1515, pp. 83–114. Andreas fault system. Proceedings of the Conference on the Tectonic
Buck, W.R., 1991. Modes of continental extension. J. Geophys. Res. 96, Problem on the San Andreas Fault System, pp. 192–205.
20161–20178. Lachenbruch, A.H., Sass, J.H., 1980. Heat flow and energetics of the San
Byerlee, J.D., 1967. Frictional characteristics of granite under high confining Andreas fault zone. J. Geophys. Res. 85, 6185–6222.
pressure. J. Geophys. Res. 72, 3639–3648. Le Pichon, X., Kreemer, C., et al., 2005. Asymmetry in elastic properties and the
Cattin, R., Martelet, G., et al., 2001. Gravity anomalies, crustal structure and evolution of large continental strike-slip faults. J. Geophys. Res. 110
thermo-mechanical support of the Himalaya of Central Nepal. Geophys. J. (B03405). doi:10.1029/2004/B003343.
Int. 147, 381–392. Lisowski, M., Savage, J.C., et al., 1991. The velocity field along the San Andreas
Chéry, J., Zoback, M.D., et al., 2001. An integrated mechanical model of the San fault in central and southern California. J. Geophys. Res. 96, 8369–8389.
Andreas fault in central and northern California. J. Geophys. Res. 106 (B10), Lowrie, A.R., Smith, R.B., 1995. Strength and rheology of the western US
22,051–22,066. Cordillera. J. Geophys. Res. 100, 17947–17963.
Cohen, S.C., Darby, D.J., 2003. Tectonic plate coupling and elastic thickness McClusky, S., Balassanian, S., et al., 2000. Global positioning system
derived from the inversion of steady state viscoelastic model using geodetic constraints on plate kinematics and dynamics in the eastern Mediterranean
data: application to southern North Island, New Zealand. J. Geophys. Res. and Caucasus. J. Geophys. Res. 105 (B3), 5695–5719.
108 (B3). doi:10.1029/2001JB001687. Meade, B.J., Hager, B.H., 2005. Block models of crustal motion in southern
D'Alessio, M.A., Johanson, I.A., et al., 2005. Slicing up the San Francisco Bay California constrained by GPS measurements. J. Geophys. Res. 110
area: block kinematics and fault slip rates from GPS-derived surface (B03403). doi:10.1029/2004JB003209.
velocities. J. Geophys. Res. 110 (B06403). doi:10.1029/2004JB003496. Meissner, R., 1986. The continental crust: a geophysical approach. Int. Geophys.
Dixon, T., Miller, M., et al., 2000. Present-day motion of the Sierra Nevada Ser. 34 (426).
block and some tectonic implications for the Basin and Range province, Melbourne, T., Helmberger, D., 2001. Mantle control of plate boundary
North American cordillera. Tectonics 19 (1), 1–24. deformation. Geophys. Res. Lett. 28 (20), 4003–4006.
Eberhart-Philips, D., Michael, A.J., 1993. Three-dimensional velocity structure, Miller, C.K., Furlong, K.P., 1988. Thermal–mechanical controls on seismicity
seismicity, and fault structure in the Parkfield Region, central California. depth distributions in the San Andreas fault zone. Geophys. Res. Lett. 15 (12),
J. Geophys. Res. 98 (B9), 15,737–15,758. 1429–1432.
 J. Chéry / Earth and Planetary Science Letters 269 (2008) 352–365 365

Nur, A., Mavko, G., 1974. Postseismic viscoelastic rebound. Science 183, 204–206. Shen-Tu, B., Holt, W., et al., 1999. Deformation kinematics in the western
Pollitz, F.F., Peltzer, G., et al., 2000. Mobility of continental mantle: evidence United States determined from Quaternary fault slip rates and recent
from postseismic geodetic observations following the 1992 Landers geodetic data. J. Geophys. Res. 104, 28927–28955.
earthquake. J. Geophys. Res. 105, 8035–8054. Sibson, R.H., 1982. Fault zone model, heat flow, and the depth distribution of
Provost, A.-S., Chéry, J., 2006. Relation between friction and fault slip rate earthquakes in the continental crust of the United States. Bull. Seismol. Soc.
across the Northern San Andreas fault system. Geol. Soc. Lond. Special Pub. Am. 72, 151–163.
253, 429–436. Sieh, K.E., Jahns, R.H., 1984. Holocene activity of the San Andreas fault at
Reilinger, R., Mc Clusky, et al., 2006. GPS constraints on continental Wallace creek, California. Geol. Soc. Amer. Bull. 95, 883–896.
deformation in the Africa–Arabia–Eurasia continental collision zone and Strehlau, J., Meissner, R., 1987. Estimation of crustal viscosities and shear
implications for the dynamics of plate interactions. J. Geophys. Res. 111. stresses from an extrapolation of experimental steady state flow data. In:
doi:10.1029/2005JB004051. Fuchs, K., Froidevaux, C. (Eds.), Composition, Structure, and Dynamics of
Rice, J.R., 1992. Fault stress states, pore pressure distributions, and the the Lithosphere–Asthenosphere System. Geodyn. Ser., vol. 16, pp. 69–87.
weakness of the San Andreas fault. In: B. E., T.-F., Wong (Eds.), Fault Townend, J., Zoback, M.D., 2000. How faulting keeps the crust strong. Geology
Mechanics and Transport Properties of Rock. Academic, San Diego, Calif., 28 (5), 399–402.
pp. 475–503. Tsenn, M.C., Carter, N.L., 1987. Upper limits of power law creep of rocks.
Roy, M., Royden, L.H., 2000. Crustal rheology and faulting at strike-slip boundaries Tectonophysics 136, 1–26.
2. Effects of lower crustal flow. J. Geophys. Res. 105 (B3), 5599–5613. Unsworth, M.J., Malin, P.E., et al., 1997. Internal structure of the San Andreas
Rybicki, K., Kasahara, K., 1977. A strike slip fault in a laterally inhomogeneous fault at Parkfield, California. Geology 25 (4), 359–362.
medium. Tectonophysics 42, 127–138. Wallace, M.J., Yin, G., et al., 2004. Inescapable slow slip rate on the Altyn Tagh
Savage, J.C., 1990. Equivalent Strike-Slip earthquake cycle in half-space and fault. Geophys. Res. Lett. 31. doi:10.1029/2004GL019724.
lithosphere–asthenosphere earth models. J. Geophys. Res. 95, 4873–4879. Wang, K., Mulder, T., et al., 1995. Case for very low coupling stress on the
Savage, J.C., Burford, R.O., 1973. Geodetic determination of relative plate Cascadia subduction fault. J. Geophys. Res. 100, 12907–12918.
motion in Central California. J. Geophys. Res. 78, 832–845. Watts, A.B., 2001. Isostasy and Flexure of the Lithosphere. Cambridge
Savage, J.C., Gan, W., et al., 2004. Strain accumulation across the Coast Ranges University Press, Cambridge.
at the latitude of San Francisco, 1994–2000. J. Geophys. Res. 109 (B03413). Williams, C.F., Grubb, F.V., et al., 2004. Heat flow in the SAFOD pilot hole and
doi:10.1029/2003JB002612. implications for the strength of the San Andreas fault. Geophys. Res. Lett. 31
Schmalzle, G., Dixon, T., et al., 2006. Strain accumulation across the Carrizo (L15S14). doi:10.1029/2003GL019352.
segment of the San Andreas fault, California: impact of laterally varying crustal Zoback, M.D., Zoback, M.L., et al., 1987. New evidence on the state of stress of
properties. J. Geophys. Res. 111 (B05403). doi:10.1029/2005/JB003843. the San Andreas fault system. Science 238, 1105–1111.
Shen, Z.K., Agnew, D.C., et al., 2003. The SCEC Crustal Motion Map, Version
3.0. . http://epicenter.usc.edu/cmm3.