Results from Prior NSF support:
Collaborative Research: Characterizing North American Upper Mantle Structure with Integrated
Inversions of USArray Surface Wave and Scattered Body Wave Data. PI’s B. Romanowicz and K. M.
Fischer. EAR-0643060 to UCB, $369,190 (06/01/07-05/31/11, including a one-year no-cost extension);
EAR-0641772 to Brown, $316,349 (06/01/07-08/31/11, including a 1.25-year no-cost extension).
This funding supported a series of integrated studies that imaged lithospheric and asthenospheric structure
beneath North America, including joint inversion of long-period teleseismic waveforms and SKS splitting
measurements for anisotropic upper mantle structure, and Sp and Ps receiver function measurements to
define major lithospheric discontinuities. This work has provided new insight on the internal structure of
the cratonic lithosphere and on differences in the character of the lithosphere-asthenosphere boundary
(LAB) between cratonic and non-cratonic regions. The main results are described in the “Background”
section of this proposal. The award partially supported a post-doc at Berkeley (Yuan) and two graduate
students (Abt and Ford), a B.Sc.-level researcher (French), and a senior thesis (Kachingwe) at Brown.
Publications supported by this grant (shown with * in references section): French et al., Geophys. Res.
Lett., 2009; Romanowicz, Science, 2009; Fischer et al., Ann. Rev. Earth Planet. Sci., 2010; Abt et al., J.
Geophys. Res., 2010; Yuan and Romanowicz, Nature, 2010a; Yuan and Romanowicz, Earth Planet. Sci.
Lett., 2010b; Yuan et al., Geophys. J. Int., 2011; Lekic et al., Science, submitted.
Background
This is a resubmission of a highly rated 2010 proposal that was not selected for funding due to
programmatic priorities. The proposal has been updated to reflect work performed in the last year
and to more fully explain our broader impacts as suggested by some reviewers.
The cratonic lithosphere
The structure of the continental lithosphere, in particular the oldest cratonic lithosphere, has long been
a subject of fascination. Cratons have remained stable since Archean times [e.g. Hoffman, 1988]. How
they were formed and how they survived destruction over timescales of billions of years remains a subject
of vigorous debate. The cratonic lithosphere presents several interesting geological and geophysical
features. Diamonds are only found in cratons and at their borders [e.g. Haggerty, 1999], seismic velocities
remain significantly higher than average down to at least 200 km depth [e.g. Gung et al., 2003], and heat
flow is low [e.g. Rudnick and Nyblade, 1999; Mareschal and Jaupart, 2004], indicating that the cratonic
lithosphere must be thick and cold. Yet, there is no observed positive geoid anomaly above cratons [e.g.
Shapiro et al., 1999; Perry et al., 2003] whereas geochemical evidence from mantle xenoliths indicates
lithosphere depletion through melt extraction [e.g. Carlson et al., 2005; Lee et al., 2011]. This has led to
the concept of tectosphere [e.g. Jordan, 1978], thick, chemically distinct cratonic mantle lithosphere that
resists destruction by subduction, owing to its neutral density and high viscosity. Nonetheless, it remains
a challenge for geodynamicists to explain why thick cratonic keels have not been progressively entrained
into the mantle by convection [e.g. Sleep, 2003; King, 2005]. The chemically depleted cratonic core may
be underlain and surrounded by a thermal, conductive boundary layer [e.g. Cooper et al., 2004; King,
2005] that acts as a buffer zone and shields the lithosphere from excessive deformation [e.g. Lenardic et
al., 2000]. The processes that formed the cratonic lithosphere are also enigmatic. Competing hypotheses
include underplating by one or more hot plumes or accretion by shallow subduction in either a continental
or arc setting [e.g. Lee, 2006]. In either case, cratonic cores were likely formed under the much different
tectonic regime of a hotter Archean mantle, which would have evolved to present day plate tectonics
some time in the late Archean, as a consequence of secular cooling [Haggerty, 1999; Griffin et al., 2003;
Carlson et al., 2005].
Seismological constraints on the structure of the cratonic lithosphere are key to understanding its
properties, origin and evolution, yet even determining the thickness of the lithosphere itself has posed a
challenge. Thermally, the intersection of the conductive geotherm with the mantle adiabat defines the
base of the lithosphere [e.g. King, 2005; Lee, 2006]. However, the thickness of cratonic roots remains
poorly defined from seismic tomography. While thicknesses in excess of 300 km have been suggested,
recent estimates, taking into account the effects of anisotropy on seismic velocities and the large negative
velocity gradient around 200 km, indicate values no larger than 200-250 km [e.g. Gung et al., 2003], in
agreement with results from xenolith and xenocryst thermobarometry [Carlson et al., 2005; Griffin et al.,
2004; Lee, 2006; Lee et al., 2011], heat flow measurements [Mareschal et al., 2004] and electrical
conductivity data [e.g. Jones et al., 2003; Evans et al., 2011].
Receiver function studies, which are more sensitive to fine-scale vertical structure than is tomography,
have in many cases failed to detect a drop in velocity that could represent the LAB beneath cratons
[Savage and Silver, 2008; Rychert and Shearer, 2009; Abt et al., 2010; Ford et al., 2010], indicating that
the velocity gradient across the cratonic LAB may not be sharp, although some receiver function studies
have inferred the presence of a cratonic LAB in certain regions [Kumar et al., 2007; Wittlinger and Farra,
2007; Snyder, 2008; Mohsen et al., 2006; Hansen et al., 2007; Hansen et al., 2009; Geissler et al., 2010;
Miller and Eaton, 2010]. On the other hand, strong Ps and Sp conversions indicating a decrease in
velocity have recently been found at shallower depths (60-120 km) under stable continental regions [Yuan
et al., 2006; Rychert and Shearer, 2009; Chen et al., 2009; Abt et al., 2010; Ford et al., 2010; Miller and
Eaton, 2010], and in other cases discontinuities at comparable depths have been characterized as
anisotropic boundaries [Bostock, 1998; Levin and Park, 2000; Saul et al., 2000; Mercier et al., 2008;
Obrebski et al., 2010a]. Comparisons with tomography indicate that these shallower cratonic
discontinuities are intra-lithospheric rather than the LAB, and they correspond to a localized drop in
velocity internal to the lithosphere seen in some surface wave models [Romanowicz, 2009; Lekic and
Romanowicz, 2010]. Such a discontinuity has also been found in long-range seismic profiles [e.g. Thybo
and Perchuc, 1997; Thybo, 2006]. Evidence for continental lithospheric layering is also well documented
from a variety of other local and regional studies [e.g. Levin et al., 1999; Snyder and Bruneton, 2007;
Chen et al., 2007; Deschamps et al., 2008; Darbyshire and Lebedev, 2009].
Variations in seismic anisotropy within the lithosphere and asthenosphere also provide important
information on cratonic lithospheric formation and preservation. In the earth’s upper mantle, seismic
anisotropy is generally attributed to lattice preferred orientation of anisotropic crystals in minerals such as
olivine and pyroxene [e.g. Nicolas and Christensen, 1987], resulting from rock deformation in past and
present mantle flow. Under continents, seismic anisotropy is often interpreted as a combination of frozenin lithospheric anisotropy from past deformation processes, shear coupling between the lithosphere and
asthenosphere, and present day flow in the asthenosphere [e.g. Park and Levin, 2002]. Anisotropy may
therefore help to constrain the geometry of mantle deformation associated with cratonic formation, as
well as the geodynamic conditions that contribute to lithospheric stability or destruction.
North America and the results of our current EarthScope research
The North American (NA) continent is in many ways an ideal target to investigate these questions
seismologically, owing to its rich tectonic history and the data collection provided by USArray and other
networks and temporary arrays. The central cratonic part of the continent is a mosaic of amalgamated
Archean blocks and Proterozoic terranes, and has remained stable since 1 Ga [e.g. Hoffman, 1989] (Fig.
1a). Successive episodes of extension, contraction, and magmatism have reworked the continent and its
margins since the formation of its stable core [e.g. Hoffman, 1989; Burchfiel et al., 1992; Thomas, 2006;
Whitmeyer and Karlstrom, 2007]. Also, subduction, extension and strike-slip deformation initiated in the
late Cenozoic is still active on the continent’s western and southwestern margin [e.g. Atwater, 1970].
Our knowledge of the 3D seismic structure of North America comes from tomographic S and P
velocity studies at a range of scales, from global [e.g. Grand et al., 1997; van der Hilst et al., 1997], to
continental [e.g. Romanowicz, 1979; van der Lee and Nolet, 1997; Godey et al., 2004; Marone et al.,
2007; Ren et al., 2007; Nettles and Dziewoński, 2008; Bedle and van der Lee, 2009], and recent higher
resolution studies focusing on the western U.S., exploiting the density of USArray stations [e.g. Sigloch et
al., 2008; Burdick et al., 2009, 2010; West et al., 2009; Obrebski et al., 2010b; Schmandt and Humphreys,
2010]. Common large-scale features of these models are thick lithosphere (~200 km) and fast velocities
under the stable part of the continent and thin (~40-120 km) lithosphere in the tectonically active west,
confirmed by recent receiver function studies [Li et al., 2007; Miller and Levander, 2009; Abt et al., 2010;
Levander et al., 2011; Lekic et al., 2011]. The transition between the active west and the craton appears
quite abrupt, closely following the Rocky Mountain Front (RMF).
Our knowledge of NA anisotropic structure has been informed by studies of core-refracted shear wave
(SKS) splitting measurements or surface wave tomographic inversions. The SKS method [e.g. Vinnik et
al., 1989; Silver and Chan, 1991] provides good lateral resolution for azimuthal anisotropy across the
continent, especially in those areas where broadband station coverage is dense, but depth resolution is
poor, and there are trade-offs between the strength of anisotropy and the thickness of the anisotropic
domain. Surface wave tomographic inversions can provide constraints on both radial anisotropy and
azimuthal anisotropy [e.g. Montagner and Nataf, 1986], and provide better vertical resolution, but they
are limited in horizontal resolution due to the long wavelength nature of the surface waves. Naturally, the
combination of the two types of data can provide better constraints, both vertically and horizontally, on
upper mantle anisotropic structure.
In previous work, we developed an approach that combines three-component surface waveforms and
SKS splitting measurements and applied it to the study of the NA continent before the USArray
Transportable Array (TA) deployment [Marone et al., 2007; Marone and Romanowicz, 2007]. In Marone
et al. [2007; from here on referred to as MR07_1], we inverted long-period teleseismic waveforms for a
continental scale 3D model of isotropic shear velocity Vs, and radial anisotropy . Lateral resolution was
on the order of 500 km for Vs and 1000 km for the radial anisotropy parameter . In Marone and
Romanowicz [2007; from here on referred to as MR07_2], we jointly inverted surface waveform data and
station averaged SKS splitting measurements for a 3D upper mantle model of azimuthal anisotropy. We
have recently demonstrated the validity of our joint inversion approach, which relies on the expressions of
Montagner et al. [2000], from the theoretical point of view [Romanowicz and Yuan, 2011]. MR07_1 and
MR07_2 identified two distinct depth domains of anisotropy that could be associated with the lithosphere
and the asthenosphere, respectively. In particular, below the LAB, the direction of the fast axis of
azimuthal anisotropy aligns with the absolute plate motion (APM), delineating an LAB with significant
depth variations (~80 to 200 km) across the continent. In the cratonic part of NA, these results agree with
several local scale studies [e.g. Li et al., 2003; Gaherty, 2004; Li et al., 2005; Snyder and Bruneton, 2007;
Chen et al., 2007; Deschamps et al., 2008; Darbyshire and Lebedev, 2009.
At the same time, we conducted Ps and Sp receiver function analysis in the northeastern U.S. and
southeastern Canada that revealed a very sharp drop in shear-wave velocity (5-10% over 5-11 km) at
depths of 90-110 km [Rychert et al., 2005, 2007]. Given that this boundary, which dips gently to the west,
fell within the broader velocity decrease from fast lithospheric lid to slow asthenosphere in existing
regional and continent-scale surface wave tomography models [e.g. van der Lee, 2002; Li et al., 2003] we
interpreted it as the LAB. These results, coupled with the observation of a strong Sp conversion from
likely LAB depths in the western U.S. using pre-EarthScope stations [Li et al., 2007], suggested that
scattered wave migration/receiver functions might be able to detect the LAB across the continent.
Initially, it seemed natural to take advantage of the potential for increased resolution offered by the
deployment of the USArray TA across NA to combine higher resolution continental-scale tomographic
inversion with constraints from ever more numerous SKS splitting measurements and from receiver
function (RF) studies. The idea was that the RFs would provide constraints on the location of the LAB
across the entire continent, so that the resulting map of LAB depths could be used as a starting point for
iterative inversions to obtain (1) higher resolution volumetric mapping of S velocity and anisotropy and (2)
adjustments to LAB depths.
However, to our surprise, one of the first results of our EarthScope project [Abt et al., 2010] was that
Sp RFs calculated for permanent broadband stations (including stations of the USArray Reference
Network) do not detect a significant negative discontinuity in cratonic regions at depths where
tomographic inversions locate the negative velocity gradient associated with the LAB, i.e. between 180250 km in NA [e.g. van der Lee and Nolet, 1997a; Marone et al., 2007; Nettles and Dziewoński, 2008;
Bedle and van der Lee, 2009] (Fig. 2 and Station ULM in Fig. 3). In contrast, a strong LAB phase is
observed in the western U.S. (Fig. 2, Station VTV in Fig. 3, and Fig. 4) and may be present in much of
the Appalachians. Our observation of pronounced LAB phases in Phanerozoic North America but a lack
of significant teleseismic scattering from the base of the North American cratonic lithosphere matches the
findings of recent studies at global [Rychert and Shearer, 2009], continental [Ford et al., 2010], and
regional [e.g. Li et al., 2007; Savage and Silver, 2008; Levander et al., 2011] scales, with the caveat that
other receiver function studies have argued for the detection of a cratonic LAB in certain regions [Kumar
et al., 2007; Wittlinger and Farra, 2007; Snyder, 2008; Mohsen et al., 2006; Hansen et al., 2007; Hansen
et al., 2009; Miller and Eaton, 2010; Geissler et al., 2010]. On the other hand, our work in North
America [Abt et al., 2010] as well as other RF studies [Yuan et al., 2006; Rychert and Shearer, 2009;
Ford et al., 2010; Chen et al., 2009] detect a velocity drop at much shallower depths (60-120 km),
suggesting the presence of a mid-lithospheric boundary [Romanowicz, 2009] perhaps related to the “8˚
discontinuity” reported in long-range seismic refraction studies [Thybo and Perchuc, 1997; Thybo 2006].
To constrain the range of possible shear velocity gradients at the respective discontinuities, the
amplitude and shapes of the phases observed on Sp RFs were modeled with synthetic seismograms [Abt
et al., 2010; Ford et al., 2010]. Due to their longer dominant periods, Sp phases typically cannot provide
resolution as sharp as that from Ps RFs. Nonetheless, modeling revealed fundamental differences in the
character of the LAB between cratonic and non-cratonic regions. For velocity drops of 10% or less (a
reasonable bound based on surface wave tomography) LAB gradients must occur over 30 km or less in
the western U.S., whereas beneath the craton they must occur over 60 km or more. In geodynamical
models for cratonic lithosphere and surrounding continental margins [King and Ritsema, 2000; Korenaga
and Jordan, 2002; Cooper et al., 2004] temperature gradients between the lithosphere and asthenosphere
occur over at least 50-70 km. We therefore conclude that while the LAB beneath the craton may be
purely thermal in origin, the LAB in the western U.S. (and possibly some portions of the easternmost
U.S.) likely requires an additional mechanism such as a more hydrated asthenosphere relative to a dry,
depleted lithosphere, or the presence of a small amount of partial melt in the asthenosphere.
As a result of the RF findings, our joint Berkeley/Brown collaboration took a different turn. Indeed,
we could not use the RF results to constrain the LAB depth in further tomographic inversions.
Nevertheless, with five more years of data collected from the recent TA deployment, as well as from
many other temporary broadband networks across the continent, we were able to significantly augment
both our waveform and SKS splitting datasets, compared to our previous studies (MR07_1 and
MR_07_2). This allowed us to achieve higher resolution and led to interesting new observations.
First and foremost, we showed that azimuthal anisotropy provides a powerful tool to detect and
document layering in the thick cratonic lithosphere [Yuan and Romanowicz, 2010a]. Indeed, beneath the
craton, we discovered that changes with depth of azimuthal anisotropy define more accurately (to within
+/-20 km) the location of the LAB than depth profiles of isotropic shear velocity (Vs) or radial anisotropy
(as defined by the parameter =(Vsh/Vsv)2). The fast axis direction of anisotropy becomes systematically
aligned with the absolute plate motion [APM, Gripp and Gordon, 2002] below the LAB, confirming our
previous findings (MR07_2). Moreover, a change in direction of the fast axis of azimuthal anisotropy at
mid-lithospheric depths clearly defines two layers within the cratonic lithosphere, separated by a
boundary with large lateral variations in depth. Layer 1 is thick under the central part of the craton and
tapers off at its boundaries with Paleozoic provinces (e.g. Fig. 1b). The thickest parts of Layer 1 are found
in regions affected by orogenies in the Archean (e.g. the Trans-Hudson orogen). Layer 1 also thins in the
Mid-continental rift zone (Fig. 5a). The lateral variations in the thickness of Layer 1 are in good
agreement with geochemical estimates from xenolith studies for the most depleted part of the craton, as
defined in terms of Mg # [Griffin et al., 2004; O'Reilly and Griffin, 2006; Yuan and Romanowicz, 2010a].
In contrast, the LAB is relatively flat under the craton (depths varying from 180 to 240 km, Fig. 5b) and
in good agreement with LAB depths predicted from geodynamical modeling, assuming that Layer 1
corresponds to the chemically depleted core of the continental lithosphere, surrounded by a thermal
“blanket” that may have formed subsequently (see Yuan and Romanowicz, 2010a for details). Azimuthal
anisotropy in Layers 1 and 2 (i.e. above 200 km depth) is primarily constrained by long-period waveform
data, whereas SKS splitting measurements help constrain the strength of anisotropy at greater depth, by
providing complementary information on the integrated effect of anisotropy across the upper mantle.
Synthetic tests confirm that the addition of SKS constraints in the inversion does not significantly change
the direction of the fast axis anywhere in the model, but stronger anisotropy is recovered below the LAB
in the craton, peaking around 270 km depth [e.g. Fig. 6; Yuan and Romanowicz, 2010a; Yuan et al., 2011].
The fast axis directions in Layer 1 generally agree with surface observations of fossil tectonic sutures,
and are coincidentally similar to the direction of APM under a large part of the craton (e.g. Figs. 1 and 5c).
This is also the direction found, as mentioned previously, in “Layer 3”, i.e. in the asthenosphere, which
provides a way to reconcile proponents of “fossil” anisotropy versus “present day flow” related azimuthal
anisotropy when interpreting SKS splitting data over NA. Both types of anisotropy are present beneath
the NA craton, and the near-surface patterns inferred from geology happen to indicate similar directions
as those found in the asthenosphere. On the other hand, Layer 2 presents a quite distinct, generally more
northerly, fast axis direction. We interpreted these results as indicative of several stages of formation of
the cratonic lithosphere and suggested that the Layer1/Layer2 mid-lithospheric boundary is likely related
to the negative velocity jump detected by RF studies and long-range seismic profile analyses. Based on
modeling of Sp RFs in both North America [Abt et al., 2010] and Australia [Ford et al., 2010], the MLD
is in some places a relatively sharp boundary (< 30 km thick) that suggests a change in composition or
fabric, as opposed to the cratonic LAB, which must have a gradual velocity gradient (> 60 km) in order to
avoid producing an Sp phase, and is therefore interpretable as a thermal boundary, i.e. the bottom of the
lithospheric thermal boundary layer. The fast axis direction in Layer 2 agrees with that expected if the
formation of the deeper cratonic lithosphere was related to accretion involving successive episodes of EW trending, N-S striking, subduction [e.g. Bostock, 1998; Mercier et al., 2008].
In the tectonically active western U.S., complex 3D patterns of isotropic velocity and anisotropy
reflect mantle dynamics associated with the rich tectonic history of the region. The alignment of the fast
axis of anisotropy with NA APM between 70-100 km depth throughout the western U.S. indicates that, to
first order, the uppermost part of the mantle is moving along with the NA plate and is not strongly
coupled to deeper mantle flow, providing some constraints on geodynamical modeling [e.g. Silver and
Holt, 2002; Becker et al., 2006]. On the other hand, at depths greater than 150 km, azimuthal anisotropy
may reflect upward and northward flow under the continent associated with the East-Pacific Rise,
constrained to the east by the western edge of the North American craton, and to the north, by the
presence of the E-W trending subduction zone. On the western side of our study region, azimuthal
anisotropy fast axes are quasi-parallel to Pacific Plate APM from ~150 km down to ~250 km depth. These
depth dependent azimuthal anisotropy patterns account for regional SKS splitting observations without
the need for any local anomalous structure under Nevada [Yuan and Romanowicz, 2010b].
In the western U.S., the LAB is shallow and we cannot associate it with a significant change in
azimuthal anisotropy direction, likely because the lithosphere is young, formed in a tectonic context
similar to the present day, and, at least to first order, the frozen-in anisotropy has a similar direction to
that marking the shear associated with the current plate motions at and below the LAB. However, the
LAB is well detected by RF studies [e.g. Li et al., 2007; Miller and Levander, 2009; Abt et al., 2010;
Levander et al., 2011; Lekic et al., 2011], and in general corresponds to a sharp drop in isotropic Vs (and
). A rapid decrease in shear modulus with depth across the LAB in the western U.S. has also been
inferred from geodetic signatures of ocean tidal loads [Ito and Simons, 2011].
Throughout the continent, directions of azimuthal anisotropy vary with depth in the lithosphere. In the
asthenosphere, azimuthal anisotropy aligns with absolute plate motion in the hotspot reference frame
[Gripp and Gordon, 2002], and manifests a maximum, stronger in the western U.S. than under the craton.
In the western U.S., from the RMF to the San Andreas Fault system and the Juan de Fuca/Gorda ridges,
this zone is confined between 70 and 150 km, decreasing in strength with depth from the top. This result
suggests that shear associated with lithosphere-asthenosphere coupling dominates mantle deformation
down to this depth in the western part of the continent. The depth extent of the zone of increased
azimuthal anisotropy below the cratonic lithosphere is not well resolved in our study, although it is
peaked around 270 km, a robust result (e.g. Fig. 6).
In radial anisotropy, >1, where  =(Vsh/Vsv)2, under the continent and its borders down to ~200 km,
with stronger  in the bordering oceanic regions. Across the continent and below 200 km, alternating
zones of weaker and stronger radial anisotropy, with predominantly <1, correlate with zones of small
lateral changes in the fast axis direction of anisotropy and with faster than average Vs below the LAB,
suggesting the presence of small scale convection with a wavelength of ~2000 km (e.g. Fig. 7).
Project Plan
The results of our recent EarthScope work lead to several interesting questions which we propose to
pursue further. A particularly intriguing finding is the prevalent layering in the cratonic lithosphere,
detected both from changes in the direction of azimuthal anisotropy and from receiver function analysis.
Within the NA craton, the boundary between Layer 1 and 2 (hereafter referred to as the MLD), as defined
from the azimuthal anisotropy analysis, varies considerably with depth, and fades out towards the
Proterozoic edges of the craton. Its depth agrees in many places with that of the boundary detected by RFs,
and ties in with results of xenolith analysis, indicating that Layer 1 may represent the most depleted,
Archean part of the continental lithosphere that is distinct from the lower cratonic lithosphere (Layer 2).
Another key result is that the LAB beneath cratonic NA (not observed in RFs) implies a much more
gradual velocity gradient (velocity drop spread over > 60 km) than the LAB beneath the western U.S.
(constrained by RFs to occur over < 30 km).
Better resolution of anisotropic layering beneath and surrounding the craton and of the MLD and
LAB will shed further light on the interpretation of Layers 1,2,3 and on the nature of their boundaries.
This will be possible in the time period of the proposed work (2012-2014). During this time, stations of
the USArray TA will be installed across the entire U.S. (and a small portion of southeastern Canada), 2
years of data collection will be complete in most of the craton, and 1-2 years of data will be available in
the Appalachians and a small corner of the craton in New York and Canada. Many open questions
remain, including the following:
1) Is the MLD as defined by a vertical gradient in azimuthal anisotropy in the tomography the same
boundary as that detected by RFs, and as the 8˚ discontinuity of long-range seismic profiles?
It is not yet clear whether the RF MLD is an anisotropic discontinuity, although the presence of
anisotropic discontinuities in the cratonic lithosphere has been suggested in some local studies [e.g.
Bostock, 1998; Levin and Park, 2000; Saul et al., 2000; Mercier et al., 2008; Snyder, 2008]. In addition,
while the RF MLD is consistent with a sharp (< 30 km) vertical boundary in some locations, in others it
appears to be more distributed in depth, perhaps with multiple layers (e.g. YKW3 and FFC, Fig. 1). The
results of the long-range seismic profiles suggest that, rather than a single discontinuity marking the top
of a low velocity layer, the 8˚ horizon [Thybo and Perchuc, 1997] marks the top of a complex zone of fine
layering, tens of km thick, manifested by strong scattering. If the proposed work shows anisotropy across
the MLD seen in RFs to be comparable to that in the waveform tomography, then they likely are the same
boundary. Alternatively, the RFs may be sensing a layering in mantle properties that differs from the
gradual vertical gradient in anisotropy in the tomography.
2) What is the nature of the cratonic MLD and how thick is the underlying low-velocity zone?
At present, the jury is still very much out. If the MLD is the boundary between two parts of the
lithosphere formed at different times and involving different processes, then the thin low velocity zone
below the MLD might be considered as a “welding” horizon between them. It could represent a zone of
kimberlite accumulation, if the strong, chemically distinct Archean Layer 1 acts as a barrier to their ascent
[e.g. Sleep, 2009], or it could be a zone of metasomatism related to the top of subducted oceanic
lithosphere welded beneath the Archean lithosphere [e.g. Chen et al., 2009], and/or it could correspond to
the right range of pressure and temperature for dehydration [Mierdel et al., 2007]. In fact, a layer of
partial melt and/or hydration has been invoked to explain the 8˚ discontinuity [Thybo and Perchuc, 1997;
Thybo, 2006] although cratonic geotherms [e.g. Griffin et al., 2004] are likely too cold to permit partial
melt below the MLD at present, except under conditions near complete water saturation [Grove et al.,
2006]. These possibilities could at least be partially distinguished by determining: (a) whether the MLD
is laterally continuous in RFs, as it appears to be in the anisotropic tomography; (b) if it is largely subhorizontal over large regions (as would be predicted by a solidus definition), or a dipping boundary
related to paleo-subduction (as suggested in the Canadian shield [e.g. Bostock, 1998; Mercier et al., 2008;
Snyder, 2008; Chen et al., 2009]); (c) the magnitude and depth extent of its isotropic velocity drop and the
orientation and depth interval of its anisotropic gradient; (d) the relationship of MLD depth and lateral
extent to xenolith constraints on mantle composition (including layering) and geotherms.
3) What is the origin of the prevailing N-S orientation of the fast axis of anisotropy in Layer 2?
Such an orientation is not obviously compatible with formation of the deep part of the continental
lithosphere by a plume impinging on the chemically distinct layer 1, as one would not expect such a
uniform direction throughout the craton, but rather a pattern that also reflects deflected upwelling from
the purported plume [e.g. Walker et al., 2007]. It is more compatible with successive accretion episodes
through subduction striking roughly north-south [e.g. Hoffman, 1989; Burchfiel et al., 1992; Thomas,
2006; Whitmeyer and Karlstrom, 2007], or with fossilized expression of drag at the bottom of the
lithosphere as it moved in a N-S direction in Mesozoic times [Deschamps et al., 2008]. However, it
remains to be tested whether the prevailing N-S direction remains stable at higher resolution, and whether
there might be a tilt of the anisotropic axis with respect to the horizontal plane.
4) How do LAB properties and origin vary between cratonic and non-cratonic regions?
Sp RFs at permanent stations imply that the LAB velocity gradient beneath the craton is gradual (> 60
km in depth) and consistent with a purely thermal origin, whereas LAB velocity gradients in the western
U.S. (and possibly portions of the Appalachians) are much sharper (< 30 km) and suggest the presence of
significant hydration or partial melt in the asthenosphere. Will this dichotomy hold up when the LAB is
more continuously sampled with data from the TA? Based on observed LAB velocity gradients, what
bounds can be placed on gradients in temperature, hydration, and melt?
5) How does the MLD under the craton relate to the LAB in Phanerozoic North America?
While an MLD within the thick cratonic lithosphere is clearly resolved, its relationship to the LAB
outside of the craton, which appears to lie at similar depths (Fig. 2) requires further study. For example,
although the Sp RFs at the permanent stations have resolved fragmentary discontinuities near the margin
of the craton, more robust and continuous RF imaging of MLD and LAB across the edge of the thick
cratonic lithosphere will be possible with the TA. Are the similar depths of the MLD and non-cratonic
LAB fortuitous? Determining whether the MLD intersects the LAB at a high angle, asymptotically
approaches the LAB, or fades in amplitude towards the edge of the craton would shed light on its origins.
6) Is there an MLD or a shallow LAB in Phanerozoic eastern North America?
Beneath the Appalachian orogen in the eastern U.S., ambiguity remains in the exact location of the
LAB and the lateral edge of the thick lithosphere. A significant negative discontinuity is observed in the
Sp RFs at permanent stations (Fig. 2). However, this discontinuity could be interpreted as a MLD, if the
LAB in the eastern U.S. is defined by the Layer 2/Layer 3 boundary in azimuthal anisotropy (e.g. Fig. 5b
and dashed line in Fig. 7 [Yuan and Romanowicz, 2010a; Yuan et al., 2011]). Or it could itself be
interpreted as the LAB, if the potential LAB depth range is defined by vertical gradients in isotropic
velocity [Abt et al., 2010] (Fig. 2) or a combination of vertical gradient in isotropic velocity and radial
anisotropy, as was used in the western U.S. by Yuan et al. [2011]. With better spatial resolution in the
eastern U.S. in both waveform tomography and RFs, the ambiguity now apparent in these different
definitions should be resolved.
Proposed Work
Sp (and Ps) migration
We propose to image upper mantle discontinuity structure, using Sp and Ps phases recorded by TA
and permanent stations in the U.S. and Canada, and also including data from past broadband temporary
arrays and from Flexible Array (FA) experiments as they become publicly available and when FA
experiment PI’s have published their initial papers. Questions to address first and foremost will be the
geographical extent, depth variation and sharpness of the MLD across the NA craton and its borders, and,
if and when it is detected, the LAB. Both Sp and Ps phases are primarily sensitive to shear-wave velocity
contrast at a given discontinuity [e.g. Rychert et al., 2007]. However, based on our experience in North
America [Abt et al., 2010; Lekic et al., 2011], Sp will likely be a more robust indicator of mantle
discontinuity structure due to possible over-printing of mantle phases by crustal reverberations in Ps RFs.
Nonetheless, because Ps receiver functions sometimes yield clear mantle discontinuities free from
apparent crustal contamination, and because their higher frequency content yields better resolution of
sharp vertical gradients in velocity, analyzing these data as a complement to Sp is worthwhile. With Sp,
we plan to minimize possible interference with unwanted teleseismic phases [Wilson et al., 2006]. We
will also experiment with incorporating SKSp, bearing in mind potential complications from other
teleseismic arrivals [Wilson et al., 2006], but the density of Sp arrivals alone should be sufficient for high
quality imaging (e.g. Fig. 4). This work will complement the research of other groups who are using Ps
[Pavlis, 2011] and Ps and Sp [Miller and Levander, 2009; Levander et al., 2011] to investigate shallow
mantle discontinuities with TA data.
Our future work will build on the efficient codes and graphical user interface for RF analysis that we
constructed for the current phase of our EarthScope project. Features include: 1) robust automated phase
picking [after Earle and Shearer, 1994]; 2) a variety of automated waveform quality checks (signal-tonoise, correlation of P and SV components, etc.) backed up by visual examination; 3) rotation into P and
SV components using the free-surface transform of Kennett [1991] and free surface Vs and Vp
determined from a grid search that best minimizes the amplitude of the parent phase (S for Sp, P for Ps)
on the daughter (P for Sp, S for Ps) component for all quality waveforms at a given station; 4) several
complementary approaches to deconvolution (extended-time multi-taper cross-correlation [Helffrich,
2006, building on Park and Levin, 2000], simultaneous frequency-domain [Bostock, 1998] and iterative
time-domain [Ligorria and Ammon, 1999]); 5) a suite of statistical tests including bootstrap estimation of
RF uncertainties; 6) mapping RFs from time to depth using a variety of velocity models; 7) calculation of
synthetic RFs for anisotropic plane-layered models based on a propagator matrix method.
Our approach for Sp and Ps imaging with TA, FA and permanent station data is to migrate RF
amplitudes into a 3D volume along the ray of the parent phase, correcting for lateral heterogeneity in
velocity, and to stack based on the geographic position of phase Fresnel zones as a function of depth; this
method is a form of common conversion point (CCP) stacking. (In portions of Canada where station
coverage is sparse, we will rely on station-based stacking.) Compared to CCP stacking, more formal
migration methods that involve forms of wave-field back propagation [e.g. Sheehan et al., 2000; Bostock
et al., 2001; Poppeliers and Pavlis, 2003a,b; Levander, 2003; Levander et al., 2006; Rondenay, 2009]
have been shown to better image dipping interfaces when data density is sufficient, but with the ~70 km
TA station spacing, they will typically yield unaliased images of mantle only at depths greater than ~140
to 200 km [Levander, 2003; Levander et al., 2006; Rondenay et al., 2005], too deep for most of the MLD
and LAB targets proposed here. Nonetheless, in addition to CCP stacking (described in more detail
below), we also plan to experiment with estimating scattering intensity by stacking along diffraction
hyperbolae from points in the model [e.g. Sheehan et al., 2000], using a pseudostation approach [e.g. Neal
and Pavlis, 1999, 2001; Poppeliers and Pavlis, 2003a,b] to reduce aliasing effects.
In on our work to date with CCP stacking of TA, dense portable array, and permanent station data in
the western U.S., we have had good success with deconvolving individual waveforms using the extendedtime multi-taper cross-correlation [Helffrich, 2006], and accounting for phase Fresnel zones in the CCP
stacking by weighting individual RF amplitudes with a normalized cubic spline function [Lekic et al.,
2011]. We correct for lateral heterogeneity using 3D models in which crustal structure beneath the
station is determined from Ps H-k stacking [Zhu and Kanamori, 2000]. For mantle structure, we have
compared CCP results assuming a variety of models [Yuan et al., 2010; Burdick et al., 2010; Schmandt
and Humphreys, 2010; Rau and Forsyth, 2011; Obrebski et al., 2011]. When applying CCP stacking in
the craton and eastern U.S., we will use mantle Vs from SAWum_NA2 [Yuan et al., 2010] and
subsequent versions and mantle Vp from Burdick et al. [2010] and subsequent versions, as well as other
future models. Uncertainties in MLD or LAB depth from bootstrap testing (two standard deviations) are
likely to be on the order of 5-20 km [Abt et al., 2010; Lekic et al., 2011]. Additional uncertainties due to
inaccuracy in corrections for lateral heterogeneity will likely be smaller; for example, no mantle
correction versus existing Vs and Vp models [Yuan et al., 2010; Burdick et al., 2009] results in MLD and
LAB depth differences of less than 6 km [Abt et al., 2010]. Images obtained with this approach in the
western U.S. reveal a clear negative Sp phase that is consistent with the LAB and whose depth variations
correlate with key tectonic boundaries. For example, in southern California the Sp LAB phase indicates
more than 30 km of lithospheric thinning beneath the surface expressions of extension in the Salton
Trough (Fig. 4a) and Inner Borderlands [Lekic et al., 2011]. Beneath the Colorado Plateau, our Sp CCP
images show an increase in lithospheric thickness to ~140 km (Fig. 4b), consistent with the findings of
Levander et al. [2011]. In much of the craton and eastern U.S., both TA data and many years of
recording from permanent station data (e.g. Fig. 2) will be available, augmented in places with FA
experiments and older portable arrays. However, to test imaging potential in those limited areas where
only TA data will be available, the Colorado Plateau 3D Sp CCP stack was re-calculated omitting all but
TA data. The resulting LAB and Moho phases are very similar to those in Fig. 4b (although the image is
somewhat noisier at depths greater than 150 km and less continuous at depths shallower than 15 km).
This comparison and the additional known potential of the permanent station data strongly suggest that
we will be able to construct accurate images of MLD and LAB discontinuities (or the lack thereof) across
the continent, enabling us to address the key questions posed in this proposal. This work will also
provide a high-resolution crustal thickness and average Vp/Vs model (based on H-k stacking of Ps at
individual stations and on Ps and Sp CCP stacking) for the long-period waveform modeling part of the
project described below.
To determine the range of velocity gradient parameters that provide acceptable fits to observed RF
phases, synthetic seismograms calculated with our existing, efficient anisotropic propagator matrix codes
will be processed in a manner identical to the data. This type of modeling will allow us to assess the
range of possible mechanisms (e.g. thermal vs. compositional gradients and/or partial melt) that could
produce the apparent LAB and MLD features and how they vary laterally (e.g. Question 4). In addition,
anisotropy produces distinctive back-azimuthal patterns in Sp and Ps RFs. For example, Fig. 8
demonstrates how anisotropy similar to that in the current long-period waveform tomography model
[Yuan et al., 2011] might appear in Sp RFs. Many permanent stations have hundreds of high-quality
waveforms, and their radial and transverse component RFs will be analyzed and modeled as a function of
back-azimuth to determine if anisotropy is possible and/or required at the MLD and LAB boundaries. In
addition, regions of the CCP stacks with high densities of conversion points will also be examined for
back-azimuthal patterns diagnostic of anisotropy. This component of the analysis will establish the
relationship between LAB and MLD features imaged by scattered waves and the LAB and MLD
boundaries determined from vertical gradients in azimuthal anisotropy in the long-period waveform
tomography (e.g. Question 1).
Anisotropic waveform modeling and tomography
The 3D radially and azimuthally anisotropic model of upper mantle shear velocity structure
(SAWum_NA2) produced during the previous funding period was developed using teleseismic
waveforms low pass filtered at 60 s, and a waveform inversion methodology, non-linear asymptotic
coupling theory (NACT, Li and Romanowicz, 1995), developed at UC Berkeley for global and continental
scale mantle tomography applications [e.g. Li and Romanowicz, 1996; Mégnin and Romanowicz, 2000;
Panning and Romanowicz, 2006; Panning et al., 2010], based on normal mode perturbation theory. This
approach has its limitations: crustal corrections are treated approximately, and lateral heterogeneity is
assumed to be weak and to vary smoothly. Moreover, large variations in discontinuity depths cannot be
accounted for.
In order to further refine the 3D structure, we can bring to bear several new features and tools: 1)
increased coverage provided by the progress of the TA across the craton and the eastern provinces; 2)
extension of the waveform inversion approach to shorter periods and the use of RegSEM, a regional
Spectral Element code, developed at IPG in Paris by Paul Cupillard [Cupillard, 2008], which has none of
the limitations of NACT. RegSEM works for large-scale regional applications, it is in spherical geometry
and it includes anisotropy, attenuation, ellipticity, perfectly-matched-layers, and non-conformal mapping
of discontinuities.; 3) inclusion of Sp constraints on Moho and MLD depths and on anisotropy at these
interfaces; 4) inclusion of Ps data (in addition to station averaged SKS splitting data) to help constrain
possible tilts of the fast axis of anisotropy.
In order to use RegSEM efficiently (i.e. with reasonable computation time), we will consider a region
that encompasses sources around north America and stations within the continent, at regional to nearteleseismic distances, and complete our current collection of three component broadband waveforms for a
couple dozen earthquakes of M>5, providing good azimuthal coverage of the region of study (e.g. Fig. 9),
using data from permanent and TA stations. Source moment tensors will be verified and updated using
regional data, as needed. We will consider a realistic crustal model, with a Moho depth constrained by
receiver functions and Vs structure constrained by surface wave dispersion data (Shapiro and Ritzwoller,
2002). We will construct a starting 3D model, modified from SAWum_NA2, to include this starting
crustal model and an explicit MLD, with depth constrained from current information on the changes in
the direction of fast axis with depth as well as Sp receiver functions. We will start with several iterations
down to 40 s, using a combination of partial derivatives computed using NACT (for distances larger than
15 deg) and PAVA (path average, i.e. standard surface wave kernels, approximate, but valid at any
distance). We will progressively incorporate more complex waveforms and move to shorter periods
(down to 20 s). In the final iterations, we will compute more accurate partial derivatives (i.e. adjoint
kernels [e.g. Tarantola, 1984; Tromp et al., 2005], which already exist for RegSEM (Cupillard et al.,
submitted). This approach is driven by the fact that SEM synthetics are very computationally intensive
and draws upon the philosophy developed in our Berkeley group for using SEM at the global scale [Lekic,
2009; Lekic and Romanowicz, 2011; French et al., in prep.].
In Fig. 10, we show a comparison of regional distance observed and synthetic seismograms, computed
using RegSEM over the North American continent for the current 3D model (SAWum_NA2), filtered
down to 40 s (Fig. 10a) and down to 20 s (Fig. 10b). The current 3D model provides a surprisingly good
fit to the long-period waveforms across the craton, down to 40 s, but, as expected, there are significant
discrepancies in both fundamental mode and overtone waveforms at shorter periods. Part of the misfit can
be explained by an inadequate crustal model, and part to more complex deeper structure.
We note that, in the latter work, as well as in our NACT inversions, misfit was defined in terms of the
point by point comparison of observed and synthetic seismograms, but we will experiment with other
misfit estimates as proposed recently in the literature [e.g. Fichtner et al., 2008, 2009], which separate
measures of misfit in phase and amplitude, thus allowing for better utilization of constraints on elastic
structure from waveform amplitudes. We note that we do not intend, at least initially and unless it
becomes evidently critical, to solve for lateral variations in attenuation, although we will test the effect of
using different existing 3D upper mantle attenuation models in the forward computations, such as
QRLW8 [Gung and Romanowicz, 2004] and QRFSI12 [Dalton et al., 2008].
At each iteration, the NA upper mantle model thus obtained will, in turn, be “fed” into the Sp
processing and serve to refine the depth and velocity contrast/gradient measurements on these boundaries
constrained by the Sp data. The new Sp results, including their implications for azimuthal anisotropy at
the MLD and other interfaces, will then be incorporated in the subsequent inversion.
For the joint waveform/shear wave splitting inversion for azimuthal anisotropy, we will increase our
existing SKS splitting data collection by performing our own splitting measurements on the new TA
stations [Menke and Levin, 2003; Yuan et al., 2008] and including those of others, as they become
available. Until now, we have inverted SKS splitting data using kernels computed under the assumption
of ray theory. For the large scales that we have been targeting, such an approximation appears sufficient.
As we move to increasingly higher resolution, we will investigate whether using finite-frequency kernels
for these measurements [e.g. Favier and Chevrot, 2003; Sieminski et al., 2008] makes a significant
difference. We will also collect splitting measurements from Ps RFs, at stations where sufficient
azimuthal coverage is available to compute the depth integrated quantities needed to jointly invert surface
waveforms and splitting data for the variations with depth of the fast axis azimuth and tilt (e.g.
Romanowicz and Yuan, 2011).
In our previous modeling efforts, we restricted interpretation to depths greater than ~50 km, due to the
lack of sensitivity of the long-period waveforms to crustal structure. Including periods down to 40 s will
allow us to make meaningful inferences on uppermost mantle structure, and to invert for perturbations to
Moho depth in the stable part of the continent, where crustal thicknesses > 30 km. On the other hand, by
including shorter period regional waveforms, we will be modeling body waves with increased sensitivity
down to transition zone depths. We expect that the resulting model will have improved resolution in the
depth range 300-500 km, i.e. into the transition zone, and in particular, will help constrain the depth
extent of the APM parallel anisotropy found under the craton below the lithosphere (Marone and
Romanowicz, 2007; Yuan and Romanowicz, 2010a), and determine whether the location of its peak
amplitude is immediately below the lithosphere, as it seems to be in the western US (e.g. Fig. 6), or
deeper into the (relatively weak) low velocity zone.
In addition to providing constraints on the volumetric and discontinuity structure in the craton and on
its borders, the SEM-based inversion will provide constraints on the sharpness and morphology of the
clear quasi-vertical boundary that lies along the Rocky Mountain Front, between the craton and the
tectonically active west. By more accurately accounting for lithospheric structure, it will also help sharpen
up our images of the cratonic asthenosphere and anisotropic structure down to transition zone depths, and
possibly help determine the presence and nature of a lower boundary to the asthenospheric layer under the
craton, which so far has been elusive.
The waveform inversion approach will likely not allow us to resolve the thickness and strength of the
low velocity zone below the MLD, however, the added constraints from receiver functions will help
remove the non-uniqueness in this type of inversion. In addition, we propose to investigate this further by
performing a set of forward modeling tests using the spectral element codes at significantly shorter
periods (~5 s) in two types of calculations:
1) We will search for a couple of adequate regional events of M>4, or sets of such events, to construct
record sections at distances between 500 and 1500 km through the craton that would provide evidence for
the presence of the 8o discontinuity in P or S waves (e.g. Thybo and Perchuć, 1997). By forward
modeling, we will adjust the fine scale structure of our 3D model around the MLD.
2) SEM modeling will also be used to enhance our understanding of the finite frequency interactions
of Sp phases with laterally varying structures, thus complementing the propagator matrix method
modeling which can only handle vertical velocity gradients. In this case, we will use our CSEM code
(Coupled-SEM, Capdeville et al., 2003) which was used in the work of Lekic and Romanowicz (2011) is
appropriate for teleseismic observations. We will test several structures beneath different stations
simultaneously. These calculations will be especially useful to consider two issues. First, in regions
where observed Sp (and possibly Ps) phases indicate abrupt lateral changes in MLD or LAB depth,
CSEM synthetics will model how these phases (at typical periods of 10-11 s for Sp) would actually
interact and resolve apparent structures. Second, where the long-period waveform tomography and other
velocity models emerging from the community resolve complex velocity structure, CSEM synthetics will
test whether the ray-based time corrections used in the scattered wave imaging are sufficient.
Interpretation
This work will produce a high resolution 3D model of radially and azimuthally anisotropic structure of
the upper mantle under the north American continent, with emphasis on the cratonic and eastern regions,
with maps of lateral variations of the MLD and LAB at a wavelength of ~200 km, and depth resolution to
less than 50 km in the lithosphere and asthenosphere, allowing unprecedented constraints on the structure
of the low velocity zones below the MLD and LAB in the stable part of the continent. Continuous CCP
images will constrain the depths of the MLD and (where possible) the LAB with a vertical accuracy of
~10-20 km, and their associated velocity gradients and anisotropic character, at smaller horizontal scales
(50-100 km). Together, these results will elucidate the relationships between the MLD, LAB, cratonic
formation, and the present-day asthenosphere.
We plan to assess the developing anisotropic velocity model, including MLD and LAB vertical velocity
gradients, in terms of temperature, composition, grain size, fabric, and melt content, using a variety of
scaling relationships [e.g. Faul and Jackson, 2005; Lee, 2003; Hacker et al., 2003; Stixrude and LithgowBertelloni, 2005; Schutt and Lesher, 2006; Karato, 2003; Hammond and Humphreys, 2000; Jackson et al.,
2006; Takei and Holtzman, 2009abc], and through interdisciplinary interactions, including experts in
xenoliths, regional geology and tectonics, and geodetic, MT and geoid/gravity models. For example, we
will collaborate with Greg Hirth (Brown) on interpreting the velocity model in terms of its implications
for mantle rheology, with a particular focus on asthenospheric viscosity beneath the craton (see letter).
Project management
Romanowicz will coordinate the Berkeley part of the project. She will contribute her experience and
insights at all stages of the long-period waveform tomography and SEM modeling. Dr. Huaiyu Yuan will
co-supervise the Berkeley graduate student in the collection of additional waveform data and SKS
splitting measurements and train him/her in the modeling tools. He will be perform the inversions and
coordinate with the Brown group for the incorporation of constraints from receiver functions. It is
possible that during the period of this project, Romanowicz will be away from Berkeley part of the time
(one semester each year). While she will be continuously available by e-mail, the co-PI’ship of Yuan will
assure that the project and supervision of the student suffers no disruption, while providing him added
opportunity for experience as an early career scientist. Fischer will supervise the scattered wave imaging.
The group will collaborate on the integration and interpretation of models and results, communicating
regularly and meeting at least twice a year at Brown or Berkeley, AGU, and the EarthScope workshop. In
addition to publishing our results and presenting them at meetings, we also plan to make our upper mantle
velocity models and discontinuity parameters openly available.
Broader Impacts
The upper mantle velocity model we will construct as well as receiver function products will provide
a continental-scale context for other researchers who are planning or interpreting other seismological
studies, including FA experiments. While we plan to interpret our results in the context of existing
mineral physics and geodynamics knowledge, our model will also be available to others for a wide range
of multi-disciplinary geological, tectonic, and geodynamical investigations of the continental lithosphere
and asthenosphere in North America and globally, in particular the fundamental question of cratonic
lithosphere formation. The proposed work will support the education and career development of early
career scientist Huaiyu Yuan, a beginning graduate student at Berkeley, graduate student Heather Ford
and a second student at Brown. Through regular presentations in group meetings and graduate seminars,
it will also contribute to the education of a group of ~15 graduate students in geophysics and 3 post-docs
at UC Berkeley and ~12 graduate students and 6 post-docs in geophysics and rock deformation at Brown.
The Berkeley Seismological Laboratory regularly hosts tours of visitors from outside the University, or
from science breadth undergraduate classes at UC Berkeley. In this framework, it will expose the public
and undergraduate students to major fundamental questions in geophysics. At Brown, Fischer will
continue to utilize this research for examples and class research projects in undergraduate and graduate
courses and for undergraduate senior theses. She and her students also regularly participate in science
outreach activities with local public schools.