3 ArticleGeochemistry
Geophysics
GeosystemsG Volume 11, Number 312 March 2010Q03008, doi:10.1029/2009GC002836
AN ELECTRONIC JOURNAL OF THE EARTH SCIENCES ISSN: 1525‐2027
Published by AGU and the Geochemical Society
Click
Here
for
Full
Article
Characterizing the 410 km discontinuity low‐velocity layer
beneath the LA RISTRA array in the North American
Southwest
John J. Jasbinsek
Physics Department, California Polytechnic State University, San Luis Obispo, California 93407,
USA (jjasbins@calpoly.edu)
Ken G. Dueker and Steven M. Hansen
Department of Geology and Geophysics, University of Wyoming, Laramie, Wyoming 82071, USA
[1] Receiver functions recorded by the 54‐station 920 km long Program for Array Seismic Studies of the
Continental Lithosphere–Incorporated Research Institutions for Seismology Colorado Plateau/Rio Grande
Rift Seismic Transect Experiment (LA RISTRA) line array display a pervasive negative polarity P to S
conversion (Pds) arrival preceding the positive polarity 410 km discontinuity arrival. These arrivals are
modeled as a low‐velocity layer atop the 410 km discontinuity (410‐LVL) and are inverted for a velocity
profile via a grid search using a five‐parameter linear gradient velocity model. Model parameter likelihood
and correlations are assessed via calculation of one‐ and two‐dimensional marginal posterior probability
distributions. The maximum likelihood model parameter values found are top velocity gradient thickness
of 0.0 km with a 4.6% (−0.22 km/s) shear velocity reduction, a 19.8 km constant velocity layer, and bottom
gradient thickness of 25.0 km with a 3.5% (+0.17 km/s) shear velocity increase. The estimated mean thick-
ness of the 410‐LVL is 32.3 km. The top gradient of the 410‐LVL is sharp within vertical resolution limits
of P to S conversion (<10 km), and the diffuse 410 km velocity gradient is consistent with hydration of the
olivine‐wadsleyite phase transformation. The 410‐LVL is interpreted as a melt layer created by the Tran-
sition Zone Water Filter model. Two secondary observations are found: (1) the 410‐LVL is absent from the
SE end of the array and (2) an intermittent negative polarity P525s arrival is observed. We speculate that
upper mantle shear velocity anomalies above the 410 km discontinuity may manifest Rayleigh‐Taylor in-
stabilities nucleated from the 410‐LVL melt layer that are being shed upward on time scales of tens of
millions of years.
Components: 9241 words, 15 figures, 2 tables.
Keywords: 410 km discontinuity; mantle transition zone; receiver functions.
Index Terms: 8124 Tectonophysics: Earth’s interior: composition and state (1212); 1038 Geochemistry: Mantle processes
(3621).
Received 9 September 2009; Revised 22 December 2009; Accepted 24 December 2009; Published 12 March 2010.
Jasbinsek, J. J., K. G. Dueker, and S. M. Hansen (2010), Characterizing the 410 km discontinuity low‐velocity layer beneath
the LA RISTRA array in the North American Southwest, Geochem. Geophys. Geosyst., 11, Q03008,
doi:10.1029/2009GC002836.
Copyright 2010 by the American Geophysical Union 1 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
1. Introduction Arizona is modeled as a 5 to 30 km thick high‐
conductivity layer atop the 410 km discontinuity
[2] The Transition Zone Water Filter (TZWF) and interpreted as evidence for a 410‐LVL
model advocates a new understanding of the genetic [Toffelmier and Tyburczy, 2007]. Globally, a 410‐
significance of the MORB and OIB geochemical LVL has been observed in eastern China
signatures found in mantle‐derived basaltic com- [Revenaugh and Sipkin, 1994], Siberia [Vinnik et al.,
position magmas [Bercovici and Karato, 2003]. The 2003], Arabia [Vinnik et al., 2004], and the SW
model suggests that the upwelling of wet wadsleyite Pacific [Thomas and Billen, 2009]. These previous
across the 410 km discontinuity may permit a results find a variable 410‐LVL thicknesses of 5–
hydrous melt layer to form atop the 410 km dis- 90 km, and a shear velocity reduction of 3–9%. It
continuity. The melt layer will sequester incompat- has been suggested that the 410‐LVL may correlate
ible elements resulting in an upwelling residuum with upwelling plume regions [Vinnik and Farra,
with a MORB like composition; thus mitigating the 2007].
need for a physically distinct upper mantle MORB
reservoir. The TZWF requires five necessary and [5] A recent Pds receiver function study using
sufficient conditions to operate: (1) wadsleyite has data recorded at 118 globally distributed stations
a higher water solubility than olivine, (2) the only infrequently detects evidence of a 410‐LVL
wadsleyite in the transition zone is >1% hydrated, [Lawrence and Shearer, 2006]. Thus, the current
(3) the mantle flow is upward, (4) the density of set of 410‐LVL observations suggests that this
the hydrous melt generated is intermediate to the layer is intermittent and spatially heterogeneous.
olivine‐dominated upper mantle assemblage and In this study, a vertical seismic resolution of 10 km
the wadsleyite‐dominated upper transition zone, is obtained using 4 s dominant period Pds arrivals.
and (5) the melt‐solid wetting angle is sufficient Stacking waveforms from the relatively high station
to create nonzero permeability to permit porous density Colorado Plateau/Rio Grande Rift Seismic
flow within the melt layer [Bercovici and Karato, Transect Experiment (LA RISTRA) array data set
2003]. Given these conditions, upwelling hydrated provides robust error estimates allowing statistical
transition zone material that crosses the wadsleyite‐ analysis of velocity model fits.
olivine phase transition at the 410 km discontinuity
is predicted to cross the wet solidus and produce a 2. Data and Methods
hydrous melt. Additionally, melt layer stability is
conditional on water content of the melt; that is, the 2.1. Array and Earthquake Sources
melt can be positively buoyant with respect to the
base of the upper mantle if the melt water content is [6] The Incorporated Research Institutions for
>7%. Hence buoyant melt‐rich diapirs may be shed Seismology–Program for Array Seismic Studies of
upward [Inoue et al., 2007; Sakamaki et al., 2006]. the Continental Lithosphere LA RISTRA array was
composed of 54 broadband seismic stations de-
[3] A simple global water filter calculation using ployed at a 17 km spacing to create a 920 km long
a background transition zone upwelling rate of linear array. The array operated for 18 months
∼1 mm/yr predicts a steady state melt layer with across New Mexico, Utah, Arizona and Texas
a thickness of 2–20 km [Bercovici and Karato, (Figure 1). Teleseismic earthquake locations used
2003]. However, a thicker melt layer could be in this study provide a good distribution of sources
formed in several ways: (1) increased wadsleyite from the NW, SE, and SW back azimuth quadrants
hydration, (2) increased mantle upflow rates, and (Figure 2). The dense station spacing and excellent
(3) the ponding of melt via lateral flow associated distribution of earthquake distances in the NW and
with 410 km discontinuity topography [Karato et SE back azimuth quadrants provide a high fold data
al., 2006]. A melt layer atop the 410 km disconti- set capable of accurately isolating Pds conversions
nuity would be manifest seismically as a low‐ from velocity discontinuities in the mantle transi-
velocity layer, herein called the “410 low‐velocity tion zone.
layer” (410‐LVL).
[4] Seismic observations of a 410‐LVL beneath 2.2. Source Deconvolution
North America are found in the Pacific Northwest
[Song et al., 2004], northern Mexico [Gao et al., [7] Receiver function calculation is performed using
2006], and the northern Rocky Mountains [Fee the Extended‐Time MultiTaper method [Helffrich,
and Dueker, 2004; Jasbinsek and Dueker, 2007]. 2006]. This method reduces spectral leakage and
In addition, magnetotelluric data beneath southern finite time series length amplitude bias by using a
2 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
azimuth quadrant is then divided into six over-
lapping 185 km wide station bins (quadrant bins)
with 50% bin sharing (Figure 1). Data culling
within each quadrant bin is performed by corre-
lating all the individual radial receiver functions
with the mean radial receiver function in each
quadrant bin and rejecting individual receiver
functions with a normalized correlation of <0.2.
The final data cull is a visual inspection of the
receiver functions that removes any harmonic or
clearly spurious traces. The final data set consists
of 1951 three‐component receiver functions. Each
quadrant bin contains 59–162 receiver functions
and the receiver functions in each bin are linearly
stacked to form a total of 18 quadrant bin stacks
(Figures 1 and 3). Receiver functions are mapped
from time to depth using the western U.S. Tectonic
North America shear wave model [Grand and
Helmberger, 1984]. The Vp/Vs ratio is fixed to the
IASP91 velocity model averages of 1.76 in the crust
and 1.81 in the mantle.
Figure 1. LA RISTRA array stations, ray sampling of
410 km discontinuity, and colored topography. Seismic
stations are denoted as black pluses. The P410s piercing 2.4. Grid Search
points for the back azimuth quadrant bins are outlined [10] To model the interfering and opposite polarity
in color. The labels A–F correspond to the six quadrant 410 km discontinuity and 410‐LVL seismic re-
stacks formed in each of the NW, SE, and SW back
azimuths. sponses, a simple five‐parameter velocity model is
used which consists of two linear velocity gradients
and a constant velocity layer. The five model
multitaper spectral estimator and overlapping spec- parameters are defined as (1) TG, the thickness of
tral estimation windows formed from orthogonal the top negative velocity gradient; (2) dVs‐TG, the
Slepian taper functions [Helffrich, 2006; Park, top negative gradient shear velocity decrement;
1987; Park and Levin, 2000]. To stabilize the (3) BG, the thickness of the bottom positive shear
spectral division in the receiver function calculation velocity gradient; (4) dVs‐BG, the bottom shear
the spectrum of preevent noise is used as the spectral
water level. After deconvolution, a second‐order
zero‐phase Butterworth band‐pass filter with cor-
ners at 3 and 30 s is applied to the receiver functions.
2.3. Data and Culling
[8] Teleseismic data recorded by the array are re-
stricted to events at 30°–100° epicentral distance
with body wave magnitudes >5.6. Earthquake
events are further culled for processing by calcu-
lating the amplitude ratio of a short‐term (2 s) tem-
poral window (STA) with respect to a long‐term
(20 s) temporal window (LTA). Events with a
STA/LTA ratio >3 are culled to yield 3690 three‐
Figure 2. Teleseismic earthquake source distribution.
component receiver functions. Earthquakes are plotted with respect to back azimuth
[9] To investigate back azimuth–dependent P s and great circle distance in degrees. The NW and SEd
seismic responses, the data are subdivided into the quadrants show an excellent distribution of sources with
three well‐sampled NW, SE, and SW back azimuth respect to distance that permits reliable moveout analy-
quadrants (Figures 1 and 2). Data from each back sis. In the SW quadrant, the event distance range istoo limited to warrant a moveout analysis.
3 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
flectivity code [Park, 1996]. Because the tangential
receiver functions have low amplitude (<2% relative
to vertical P) and no coherence with respect to the
radial components, (Figures 3d–3f) this choice of
isotropic calculations is reasonable. The mean ray
parameter value of the data set is about 0.06 s/km
and is used in the synthetic calculations. To account
for anelastic effects, the Futterman attenuation
operator [Futterman, 1962] is applied to the syn-
thetic data. A t*Pds value of 1.5 s is used assuming a
t*P410s value of 2 s and a t*P410p value of 0.5 s [Liu,
2003]. Application of this attenuation operator
results in a small amplitude decrease of the main
pulse and a slight broadening of pulse widths.
[12] Since we are interested in modeling waveform
shapes and not the absolute timing of the wave-
forms, the synthetics are cross correlated with the
waveform stacks to find the optimal temporal
alignment. The lag time of the peak of the cross
correlation function is used to shift the observed
and synthetic waveforms into relative alignment.
Both the L2 and cross‐correlation norms have been
used as a misfit function, but these two norms show
little difference in the posterior probability density
functions. Thus, the L2 norm is used because it
permits Gaussian statistics to be used and posterior
probability density functions may be calculated.
2.5. Marginal Posterior Probability Density
Function Integration
Figure 3. (a–c) Radial and (d–f) tangential receiver [13] A Gaussian likelihood function is used to
functions from NW, SE, and SW back azimuths. Quad- quantify the fit of each model to the data:
rant bins A–F are shown in Figure 1. The number of 1
receiver functions in each bin is indicated at the bottom LðmÞ ¼ k exp ðd gðmÞÞTC1D ðd gðmÞÞ ;
of each trace. The signal that is greater than one standard 2
deviation is shaded red/blue, and the standard deviation
of the signal is indicated as thin black lines. The am- where d is the observed receiver function vector, m
plitude scale symbol denotes a 10% variation relative to is the double gradient slab model vector, g(m) is the
the vertical P wave amplitude. synthetic receiver function for a given model m, C
−1
D
is the inverse data covariance matrix [Sambridge,
1999], and k is an arbitrary scaling constant, taken
velocity gradient increase; and (5) ST, the thickness to equal 1. The probability of a modelm given a data
of a constant velocity layer bounded above and vector d is calculated as P(m∣d) = kr(m)L(m∣d)
below by the two velocity gradients. The dVs‐BG where k is a second arbitrary scaling factor and r(m)
parameter is defined as the velocity increase in the is the a priori probability distribution of the model
bottom gradient after recovery of the top gradient parameters. In this study all models are considered
velocity decrement (Figure 4a). equally likely a priori, i.e., r(m) = 1; hence, the
[11] To assess the fit of the double‐gradient slab probability and likelihood functions are simply
model (DGS) with respect to our stacked wave- rescaled versions of each other and the probability
forms, a five dimensional grid search is performed function P(m) is replaced by the likelihood func-
resulting in the calculation of 116,160 synthetic tion L(m).
seismograms (Table 1). Synthetic receiver func- [14] To visualize the five‐dimensional error volume
tions are calculated using an isotropic velocity re- defined by L(m), the 1‐ and 2‐D marginal proba-
4 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
Figure 4. Double gradient slab (DGS) model parameterization. (a) The solid black line is the IASPI91 shear velocity
model, and the dashed red lines indicate a trial DGS velocity model. The five model parameters are as follows: TG,
top gradient thickness (km); ST, constant velocity slab thickness (km); BG, bottom gradient thickness (km); dVs‐TG,
shear wave velocity decrease for the top gradient (km/s); and dVs‐BG, shear wave velocity increase in the bottom
gradient (km/s). The value of dVs‐BG is the velocity increase after the top gradient velocity decrement is recovered.
(b) IASPI shear velocity model and maximum likelihood DGS velocity model for the Global Waveform Stack (GWS)
(Figure 6). The velocity model is characterized by a sharp top velocity gradient and diffuse bottom velocity gradient.
bility distributions are computed for each model pairs of model parameters. To test our calculation
parameter and pairs of model parameters (Figure 5). of the marginal probability distributions, synthetic
The 1‐D marginal probability distribution function data with an additive normally distributed noise
for the ith model parameter is defined as with a standard error equal to that measured in the
Z Z Y culled data set are created. Our calculation ofn
M1ðmiÞ ¼ PðmÞ dmk ; posterior marginal probability density functions
k¼1 recovers near‐exact parameter values of the syn-
k 6¼i thetic velocity model when applied to this synthetic
data.
where n is the number of DGS model parameters
(five). The 2‐D marginal probability distributions
between the ith and jth model parameter are simi- 3. Results
larly defined as
Z Z [16] To characterize the trade‐off between spatialYn bin size and observed back azimuth–dependent
M2ðmi;mjÞ ¼ PðmÞ dmk :
k¼1
k 6¼i;j Table 1. Parameter Value Ranges for the Double‐Gradient
Slab Model
[15] The 1‐D marginal probability distributions in- Parameter Minimum Increment Maximum
dicate how well a model parameter is resolved: a TG (km) 0 5 45
compact unimodal distribution indicates a well‐ BG (km) 0 5 35
resolved model parameter. The 2‐D marginal prob- ST (km) 0 5 55
ability distributions display the correlation between dVs‐TG (km/s) −0.80 0.08 0
dVs‐BG (km/s) 0 0.17 1.7
5 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
Figure 5. Global waveform stack 1‐ and 2‐D marginal probability distributions for the five double gradient slab
model parameters. The model parameters are denoted as follows: TG, top gradient; BG, bottom gradient; ST, slab
thickness; dVs‐TG, top shear velocity step; dVs‐BG, bottom shear velocity step. The 1‐D distributions are shown
along the diagonal of the upper half “matrix,” and the 2‐D distributions are the “off‐diagonal matrix” elements.
The 1‐D distributions are unimodal and compact, indicating well‐resolved model parameters. The 2‐D distributions
are contoured at the 99% probability level and show that the model parameters are uncorrelated.
Pds responses, three different spatial scale stacks terization. The DGS maximum likelihood parame-
are presented: the stack of all 1951 receiver ter values are: a top gradient thickness of 0 km; a
functions termed the global waveform stack, the top gradient shear velocity decrement of 0.22 km/s;
stacks of the receiver functions by back azimuth a bottom gradient thickness of 25 km; a bottom
quadrant bin termed quadrant stacks, and a com- gradient shear velocity increase of 0.17 km/s; and a
mon conversion point (CCP) image [Dueker and constant velocity layer thickness of 19.8 km
Sheehan, 1997] that uses 80 km wide bins. Each (Figure 4b). The 2‐D marginal posteriori proba-
of the three analyses shows the presence of a bility distributions show tightly peaked functions
negative polarity Pds arrival interfering with the indicative of little trade‐off between the five model
P410s arrival. parameters (Figure 5). Thus, the DGS model pa-
rameterization provides a very good fit to the mean
3.1. Global Waveform Stack stacked waveform (Figure 6b) with a reduced chi
square value of 1.3. The thickness of the 410‐LVL is
[17] The global waveform stack of all 1951 radial estimated as the constant velocity layer thickness
receiver functions shows a clear negative polarity plus half the sum of the top and bottom gradient
Pds arrival (410‐LVL arrival) preceding and inter- thicknesses: i.e., 32.3 km (Figure 4b). With respect
fering with the positive polarity P410s arrival to the IASPI91 velocity model, the top gradient
(Figure 6a). The 1‐D marginal probability dis- shear velocity decrement is 4.6% (−0.22 km/s)
tributions for the DGS model are highly peaked and the bottom shear velocity increment is 3.5%
(Figure 5), demonstrating that the DGS model (0.17 km/s).
(Figure 4) is an adequate velocity model parame-
6 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
stacks labeled as A–F (Figures 1 and 3). Inspection
of the radial component quadrant stacks shows
that the 410 and 660 km discontinuity arrivals are
accurately isolated (Figure 3). The waveform of
the 660 km discontinuity arrival is coherent across
the quadrant bins and the peak amplitude of the
arrival is consistent with the predicted IASP91
P660s arrival amplitude of 3% relative to vertical
P. The 410 km discontinuity arrival amplitude and
pulse width is more variable and are discussed
below. In addition, the SE quadrant stacks for
bins C–E (Figure 3b) show a set of positive arri-
vals at 440 km that are underlain by an irregular
set of small negative polarity arrivals. However,
these secondary arrivals are not observed in the
NW and SW back azimuth quadrants (Figures 3a
and 3c).
[19] The tangential component quadrant stacks
contain low amplitudes of <2% with respect to the
direct P wave, consistent with the absence of sig-
nificant anisotropic velocity contrasts or dipping
layers (Figures 3d–3f). Where tangential energy is
observed, no coherence is observed with respect
to the radial component quadrant stacks. This
observation suggests that the effects of velocity
anisotropy and dipping layers may be ignored in
our analysis.
[20] Above the 410 km discontinuity arrival, nega-
tive polarity P375s (410‐LVL) arrivals are observed
on 14 out of the 18 radial receiver function quadrant
stacks. Generally, the 410‐LVL arrivals have equal
or greater amplitude with respect to the 410 km
discontinuity arrival. The most notable spatial vari-
ation of the 410‐LVL arrival is its disappearance
at the SE end of the array beneath SW Texas and
Figure 6. Global waveform stack analysis. (a) Global SE New Mexico (Figures 3b, 3c, 7, and 8).
waveform stack (shaded red/blue) and standard devia-
tion error lines. (b) Windowed global waveform stack [21] From the quadrant stacks, the measured depth
410 and 410‐LVL and model fit. The maximum likeli- range of the 410 km discontinuity is 389–406 km
hood model values (i.e., the peaks of the marginal prob- (17 km variation) with a mean of 403 km and the
ability distributions in Figure 5) are used to calculate the depth range of the 660 km discontinuity is 673–
synthetic seismogram denoted as the thickened black 679 km (6 km variation) with a mean of 676 km
line. The reduced chi‐square value of the synthetic (Table 2). The errors associated with these depth
model is 1.3. (c) Moveout/phasing characteristics of the estimates are assessed via bootstrapping [Efron
410‐LVL, 410, and 660 km discontinuity. The unity and Tibshirani, 1994]. This error analysis finds
slope gray line shows the line of predicted Pds moveout the standard deviation of the discontinuity depths
at which the amplitude of P to S conversions should be is generally <3 km. However, this error analysis
maximal. The results demonstrate that the 410‐LVL, only quantifies the error in isolating the peak arrival
410, and 660 km discontinuities display the correct Pds
moveout. time of the discontinuity arrivals. The uncertainty
in the absolute discontinuity depths are contingent
upon the accuracy of the assumed P and S wave
3.2. Quadrant Stacks velocity model used to migrate the receiver func-
[18] In each back azimuth quadrant bin, the receiver tion from time to depth. Using reasonable varia-
function data are linearly stacked to form quadrant tions in our P and S wave velocity model [Bedle
7 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
Figure 7. Pds common conversion point image and a shear wave tomogram produced from the LA RISTRA array
[Sine et al., 2008]. Crustal depth is indicated by the grey shaded region [Wilson and Aster, 2005]. The down arrows
indicate regions of potential downwelling due to lithospheric edge delamination [Song and Helmberger, 2007], and
the up arrows indicate regions of potential upwelling associated with low velocity mantle anomalies. Multiple arrivals
beneath the 660 km discontinuity at the SE edge of the image are not interpreted due to the reduced sampling at the
edge of the image.
and van der Lee, 2009; Burdick et al., 2008, 2009; and the simplest 410‐LVL waveforms (Figure 3a).
Sine et al., 2008] the absolute discontinuity depths An overlay of the maximum likelihood double‐
are estimated to be accurate to within ±5 km. gradient slab model synthetic waveforms generally
Thus the 410 and 660 km discontinuities are 15 and fit within one standard deviation of the stacked
16 km shallower and deeper, respectively, than waveforms with reduced chi‐square values of 0.5–
averages (418 km and 660 km) beneath the North 8.2 (Figure 9). The double gradient slab 1‐D mar-
American southwest [Flanagan and Shearer, ginal probability functions for the NW quadrant
1998b]. stack bins (Figure 10) show that the velocity steps
for the top and bottom velocity gradients are very
[22] Because the magnitude of velocity anomalies consistent (−0.37 km/s and 0 km/s, respectively).
in the transition zone are generally much smaller For the top velocity gradient thickness, four of the
with respect to the upper 200 km, the mean 271 km six 1‐D marginal probability distributions have a
transition zone thickness beneath the array is con- peak probability density at 0 km thickness; whereas
sidered a robust estimate. This estimate is 11.5% for the bottom gradient thickness, the six marginal
thicker than the global average of 243 km [Gu probability distributions are centered near 22 km.
and Dziewonski, 2002], suggesting a cold and/or The constant velocity slab (ST) marginal proba-
hydrated transition zone [Bolfan‐Casanova et al., bility distributions are the most variable, indicating
2006; Courtier and Revenaugh, 2006; Frost and lateral variations in the thickness of the 410‐LVL.
Dolejs, 2007; Hirschmann et al., 2005; Litasov
et al., 2006; Wood, 1995]. [24] The SW quadrant stack bins E and F (Figure 3c)
lack a 410‐LVL arrival and hence provide a simpler
[23] To quantify variations in velocity character- seismic response with which to constrain the 410 km
istics of the 410‐LVL within the array, the NW discontinuity velocity step. Modeling of the wave-
quadrant stacks are modeled with the double‐ forms from these two bins finds a bottom gradient
gradient slab model (Figure 4 and Table 1). The thickness of 16.7 and 30.0 km and a shear velocity
NW quadrant stacks are used because they contain increment of 0.51 and 0.49 km/s, respectively
the most receiver functions per bin (mean of 145) (Figure 11). As expected due to the lack of a
8 of 17
Geochemistry
Geophysics 3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836
Geosystems G
Figure 8. Overlay of quadrant stack traces and S wave tomogram of Sine et al. [2008]. (a) NW quadrant stack
overlay (see also Figure 3a). (b) SE quadrant stack overlay (see also Figure 3b). The scale bar displays the tomo-
graphic S wave velocity perturbations. Receiver functions (black lines) are oriented such that positive polarity is to the
right. The quadrant stack waveforms are traced to depth from the middle of each bin at the surface (Figure 1) using an
average ray parameter of 0.06 s/km. Quadrant stack waveform amplitudes are muted above 200 km depth. Labels
denote the following: M, Moho; 410, 410 km discontinuity; RGR, Rio Grande Rift; FC, Four Corners region. Labels 1
and 2 indicate regions of low S velocity above the 410 km depth, and grey arrows indicate potential flow direction.
Label i is interpreted in previous studies as a Farallon slab fragment, and label ii is a possible delaminated volume of
lithosphere.
410‐LVL arrival, the double‐gradient slab model- and 410‐LVL discontinuities (Figure 7). In addi-
ing recovers a model with no velocity decrement at tion, an intermittent negative polarity arrival is
the top. Yet, the estimated 410 km discontinuity found at 520–560 km depth. The 660 km discon-
shear velocity increase is twice the IASP91 pre- tinuity arrival is coherently imaged with mean
dicted value of 0.23 km/s. These large amplitude amplitude of 3% with respect to the P component
P410s arrivals are likely focused arrivals from local amplitude beneath the stations where data fold
410 km discontinuity topography and thickness is good. The 660 km discontinuity image finds
variations and/or near surface amplification effects no evidence of significant topography. The most
[van der Lee et al., 1994]. notable change in the 410 km discontinuity signal
is found beneath SE New Mexico and SW Texas;
here, the CCP image finds that the 410‐LVL arrival
3.3. Common Conversion Point Image is absent. This spatial variation in the 410‐LVL
[25] A common conversion point (CCP) image of arrival is also observed in the SE and SW quadrant
the 1951 receiver functions images the 410 and 660 stacks (Figures 3b and 3c).
9 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
Table 2. Depths and Errors for the 410 and 660 Discontinuities
Quadrant Stack d410 (km) d410, s (km) d660 (km) d660, s (km)
NW_A 405.0 3.0 673.8 1.7
NW_B 401.6 2.9 676.5 1.9
NW_C 401.1 4.1 679.0 1.5
NW_D 400.4 2.6 679.3 1.2
NW_E 401.3 2.9 676.8 2.2
NW_F 406.3 2.5 676.7 2.6
SE_A 403.0 3.9 677.9 2.1
SE_B 404.0 6.3 678.5 2.9
SE_C 405.0 5.1 677.8 4.6
SE_D 403.3 3.5 675.5 5.9
SE_E 405.9 3.1 674.5 2.5
SE_F 397.0 4.5 676.0 9.6
SW_A 402.5 4.1 675.5 2.8
SW_B 404.6 2.0 675.6 2.0
SW_C 404.2 1.5 674.1 3.3
SW_D 404.5 1.6 675.2 4.1
SW_E 406.0 3.0 674.7 2.7
SW_F 403.4 2.6 674.1 3.2
Mean 403.3 3.3 676.2 3.2
[26] Noteworthy is that a previous study of the LA for selected wave directions. However, the tan-
RISTRA 410 and 660 km discontinuities only found gential receiver function stacks have small ampli-
weakly intermittent 410‐LVL arrivals [Wilson et al., tudes and no spatial coherency consistent with
2005b; Wilson and Aster, 2005] with much less velocity anisotropy (Figure 3). Therefore, velocity
lateral coherence than the images presented herein. anisotropy is rejected as a cause of the 410‐LVL.
This difference in the lateral coherence of the Second, a subsolidus chemical anomaly with respect
410‐LVL arrival is attributed to different culling to an assumed pyrolitic mantle, such as accumulated
of the data and the larger bin sizes used in our oceanic crust separated from its underlying mantle
receiver function quadrant stacks. lithosphere [Lee and Chen, 2007] could create a
low‐velocity anomaly. Yet, tomographic images of
the upper mantle and transition zone in the study
3.4. Negative Polarity P525s Arrivals area do not support the presence of oceanic crust
[27] Our global waveform stack finds a 2% nega- atop the transition zone [Bedle and van der Lee,
tive polarity amplitude arrival at about 525 km 2009; Burdick et al., 2008, 2009; Sine et al.,
depth. This arrival is prominent in the SW quadrant 2008]. Therefore, the 410‐LVL is interpreted as a
stacks and intermittently in the NW and SE quadrant hydrous melt layer created by the Transition Zone
stacks at 520–550 km depth (Figure 3). In addition, Water Filter [Bercovici and Karato, 2003].
the CCP image intermittently images this arrival in
the 530–550 km range (Figure 7). Due to the lower
signal‐to‐noise ratio of this arrival, the moveout
characteristics are not as well resolved as the
410‐LVL, P410s and P660s arrivals (Figures 6c and
A1). The proximity of the observed 525 km arrival
to the nominal depth of the wadsleyite‐ringwoodite
phase transformation (520 km discontinuity) is
discussed below.
Figure 9. Overlay of maximum likelihood double gra-
4. Discussion dient slab model synthetic and observed waveforms for
the NW event back azimuth quadrant. Numbers adjacent
[28] Several seismic velocity effects could explain to each trace are the reduced chi‐square values for model
the origin of the 410‐LVL. First, velocity anisotropy fit to data. The gray scale bar represents 10% amplitude
due to lattice preferred orientation just above the variation relative to vertical P. The maximum likelihood
model fits the peaks and troughs of the observed data
410 km discontinuity could produce low velocities within the one‐sigma error bars.
10 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
Figure 10. Overlay of the NW quadrant stack A–F binned waveform 1‐D marginal probability distributions for the
double gradient slab model (Figure 4a). The model parameters are denoted as follows: TG, top gradient; BG, bottom
gradient; ST, slab thickness; dVs‐TG, top shear velocity step; dVs‐BG, bottom shear velocity step. The colored lines
(A–F) correspond to the stacking bins (A–F) shown in Figure 1. The peaks of the 1‐D marginal probability distri-
bution functions define the maximum likelihood parameter values for the double gradient slab velocity models.
4.1. Transition Zone Thickness hydrated upper and lower mantle transition zone.
A possible mechanism for transporting water into
[29] The olivine‐wadsleyite transformation (410 km
the transition zone is via subducting slabs. Dense
velocity discontinuity) is predicted to occur at lower
hydrous magnesium silicates found in subducting
pressure and over a broader pressure interval un-
slabs transform into superhydrous phase B and
der hydrous conditions [Frost and Dolejs, 2007;
phase D at transition zone pressures [Irifune et al.,
Hirschmann et al., 2005, 2006; Litasov et al.,
1998; Kohlstedt et al., 1996; Shieh et al., 1998];
2006; Smyth and Frost, 2002; Wood, 1995]. Al-
though results vary based on experimental method,
starting material composition and water content, the
results of Litasov et al. [2006] are an illustrative
example: in hydrous pyrolite containing ∼3 wt %
H2O the olivine‐wadsleyite transformation is esti-
mated to occur at 0.5 GPa (15 km) lower pressure,
with a significant broadening of the transformation
pressure interval to 1.2 GPa (36 km). This is broadly
consistent with our finding of 16 and 30 km wide
olivine‐wadsleyite transition thicknesses in quad-
rant stacks SW‐E and SW‐F, respectively (Figures 3
and 11), and 410 km discontinuity depths decreased
by 12 and 14.6 km from the 418 km average in the
North American southwest [Flanagan and Shearer,
1998a]. Conversely, the 660 km discontinuity is
predicted to occur at higher pressure in a hydrous
versus anhydrous environment [Bolfan‐Casanova et
al., 2003, 2006]. One study estimates the 660 km
discontinuity onset pressure is increased by 0.6– Figure 11. Maximum likelihood models for SW‐E and
SW‐F quadrant bin stacks that do not have a significant
1 GPa (18–30 km) in hydrous pyrolite [Litasov et al., 410‐LVL arrival. The black scale bar represents 10%
2006], consistent with our mean observed 660 km amplitude relative to vertical P. For the SW‐E and
discontinuity depth of 676.2 km (Table 2). Thus, the SW‐F bins, bottom velocity gradient thicknesses of
larger than average transition zone thickness found 16.7 km and 30.0 km and shear velocity increases of
here may be explained by perturbations of both 0.51 km/s and 0.49 km/s, respectively, are found by
nominal 410 and 660 km discontinuity depths via a the double gradient slab model fitting.
11 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
thus, these phases may provide the necessary respect to the 410 km phase transition: water sol-
transport mechanism. ubility may be locally greater in the shallower
wadsleyite region (up to 3.2 wt% H O) than in the
[30] Temperature perturbations also modulate the 2
deeper ringwoodite region (up to 2.7 wt%) [Bolfan‐
onset pressure of the 410 and 660 km discontinuities
Casanova et al., 2006; Inoue et al., 1998; Kohlstedt
[Bolfan‐Casanova et al., 2006; Frost and Dolejs,
et al., 1996; Ohtani et al., 2000]. Thus, a down-
2007; Hirschmann, 2006; Hirschmann et al.,
ward flux of sufficiently hydrated wadsleyite across
2005; Katsura et al., 2004], with colder tempera-
the 520 km phase transition would require re-
tures decreasing/increasing the pressures at which
partitioning of the water [Karato et al., 2006]. The
the 410/660 km discontinuities occur. Applying a
exsolved water may sufficiently lower the solidus of
Clapeyron slope of 3 MPa/K for the 410 km dis-
ringwoodite to produce either hydrous melt or free
continuity, the maximum temperature decrease
water vapor [Karato et al., 2006]. If true, then a
required to explain the up to 14.6 km shallower
negative polarity P s arrival could be produced.
than normal 410 km discontinuity depths in the d
SW‐E and SW‐F quadrant stacks is 163°C. Sim- [33] An important difference between the 410 km
ilarly, a temperature decrease of 600–1200°C is discontinuity water filter and a potential 520 km
required to explain the mean 660 km discontinuity discontinuity water filter is that no density contrast
with a Clapeyron slope of −1 to −0.5 MPa/K. exists for the melt to gravitationally perch upon.
Tomographic images of the mantle beneath North Thus, assuming the melt produced is positively
America are consistent with the presence of slab buoyant, it would percolate upward. Given the
fragments in the transition zone [Bedle and van der uncertainties in the potential for hydrous melting
Lee, 2009; Burdick et al., 2008; Roth et al., 2008; across the wadsleyite‐ringwoodite phase boundary
Sine et al., 2008], which could explain lower than and melt density about the 520 km discontinuity,
normal mantle transition zone temperatures. How- this interpretation is speculative. We note also that
ever, detailed temperature perturbations in the negative polarity Pds arrivals have been observed at
transition zone beneath the linear 2‐D LA RISTRA 570–600 km depths in the transition zone beneath
array are not resolvable. southern Africa [Shen and Blum, 2003]. These
arrivals were modeled as a 2.2% shear velocity re-
4.2. Negative Polarity P s Arrivals duction and interpreted as evidence for accumulated525 oceanic crust above the 660 km discontinuity.
[31] Negative polarity P525s arrivals (Figures 3, 6a,
and 7) cannot be explained by any known solid
state phase transformations in the transition zone. In 4.3. Buoyant Upwelling From the 410 Low‐
particular, the 520 km discontinuity, which is gen- Velocity Layer
erally associated with the wadsleyite‐ringwoodite [34] A dynamical question with respect to the
phase transformation, is a positive velocity step with hydrous melt layer atop the 410 km discontinuity is
respect to increasing depth [Helffrich and Wood, its gravitational stability [Leahy and Bercovici,
2001] thus creating a positive polarity P‐S conver- 2007]. To first order the gravitational stability of
sion. The dominant mantle transition zone minerals, a 410 melt layer depends on the melt layer thick-
wadsleyite and ringwoodite, have higher water sol- ness and the hydrous melt density contrast with
ubility than the dominant upper (olivine) and lower respect to the upper mantle [Leahy and Bercovici,
(perovskite + ferropericlase) mantle minerals [Bolfan‐ 2007; Youngs and Bercovici, 2009]. The hydrous
Casanova et al., 2002, 2003;Hirschmann et al., 2005; melt density and viscosity are strong functions of
Murakami et al., 2002]. These solubility contrasts water content. Experimental results find that in a
may cause the mantle transition zone to act as a pyrolitic assemblage melt production at 1400°C
mantle water reservoir [Karato et al., 2006]. with 6.7 wt% water forms a gravitationally stable
[ ] Similar to the olivine‐wadsleyite phase tran- hydrous melt perched atop the 410 km disconti-32
sition at the 410 km discontinuity and the ring- nuity density gradient [Matsukage et al., 2005;
woodite‐perovskite phase transition at the 660 km Sakamaki et al., 2006]. However, a gravitationally
discontinuity, the wadsleyite‐ringwoodite phase unstable layer is produced at temperatures of 1400°
transition demarks a change in water solubility C and water content of 10 wt% in the melt. Thus,
[Bolfan‐Casanova et al., 2006]. However, the di- the development of a Rayleigh‐Taylor instability
rection of the decrease in water solubility across the from a sufficiently thick and water‐rich 410 melt
520 km phase transition is potentially opposite with layer is plausible [Inoue et al., 2007].
12 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
[35] The temporal evolution of a Rayleigh‐Taylor Mexico (Figures 3, 7, and 8). Assuming that the
instability from the 410‐LVL would depend on the fast material in the transition zone is delaminated
volume of positively buoyant melt layer that is lithosphere [Song and Helmberger, 2007], then the
available to feed the instability. Gravitational seg- 410 melt layer, assuming it had been previously
regation of the hydrous melt layer would segregate developed, has either (1) advected downward be-
the most water rich, hence buoyant, melt to the top neath the sinking lithospheric volume or (2) flowed
of the melt layer. If sufficient volume is available, laterally due to the pressure associated with the
an upwelling plume‐like diapir may be expected as sinking lithospheric volume as it transited the
buoyant melt forces the growing instability [Youngs 410 km discontinuity.
and Bercovici, 2009]. Eventually, a growing insta-
[39] Buoyantly ascending hydrous melt diapirs
bility may locally exhaust available buoyant melt
(“water pipes”) may be an explanation for the low‐
volume leading to a detachment of the diapir from
velocity anomalies above 410 km discontinuity
the 410‐LVL.
beneath the Rio Grande Rift and the Four‐Corners
[36] Assuming the top 20 km of the 32.3 km thick area (Figure 8) [Wilson et al., 2005a]. The impact
410‐LVL found by our analysis is gravitationally of buoyant hydrous melt diapirs on 10 million year
unstable, the dominant wavelength of the R–T in- time ascent scales beneath the Rio Grande Rift
stability is predicted to be ∼50 km [Turcotte and could provide potential gravitational energy to
Schubert, 2002; Youngs and Bercovici, 2009]. A promote extension of the lithosphere [Jones et al.,
50 km diameter spherical diapir volume would 1996; Wilson et al., 2005a]. Further high‐resolution
require advection of a 20 km thick layer over a seismic imaging of the upper mantle beneath this
3250 km2 area (57 km wide square). Assuming a region may allow this “water pipe” hypothesis to be
10% melt solid density contrast and a 1020Pa s tested.
mantle viscosity, a Stokes Flow estimate [Turcotte
and Schubert, 2002] of melt volume transit time 5. Conclusions
to the base of the lithosphere is on the order of
10 million years.
[40] This study has constrained the thickness and
[37] A S wave tomogram constructed from the LA velocity characteristics of a low‐velocity layer atop
RISTRA array data [Sine et al., 2008] (Figures 7 the 410 km discontinuity (410‐LVL) beneath the
and 8) defines many low‐velocity anomalies in southern Rocky Mountains with a dense array data
the upper mantle and transition zone. The first set set. The 410‐LVL is found to be 32.3 km thick
of velocity anomalies is centered at −109° longi- with a sharp top velocity gradient with a mean
tude and originates near the 410 km discontinuity 4.6% shear velocity reduction. These results are
(labeled 1 in Figure 8a). These velocity anomalies interpreted as a manifestation of the Transition
reside near the Four‐Corners and Jemez caldera Zone Water Filter model that has created a layer of
just west of the Rio Grande Rift. The second set of hydrous melt atop the 410 km discontinuity. The
velocity anomalies above 410 km depth reside at receiver function data also isolates a negative
−104° longitude (labeled 2 in Figure 8a) [Sine et polarity Pds conversion at 525 km. We speculate
al., 2008]. The maximum shear velocity decrease this results from a hydrous melting process similar
in region 1 is ∼3%, corresponding to a 1.4% melt to the TZWF model. Comparison of the 410‐LVL
porosity using standard upper mantle melt velocity results with S wave tomography from the LA
scaling relations [Kreutzmann et al., 2004]. Using RISTRA Array [Sine et al., 2008] suggests the
the same scaling relations, the 4.6% shear velocity 410‐LVL may be shedding buoyant hydrous melt
reduction at the 410‐LVL corresponds to 2.1% diapirs that ascend upward to impact the base of the
melt porosity. Therefore, a rising diapir from the lithosphere.
410‐LVL could possess sufficient melt porosity to
account for the observed shear velocity reductions. Appendix A
We speculate that these low‐velocity anomalies in
the upper mantle may be ascending diapirs shed A1. Moveout and Resampling Analysis
from the 410‐LVL melt layer.
[41] The P410s, P660s and P410‐LVLs arrivals from[38] Finally, it is remarkable that the 410‐LVL is
the NW and SE back azimuth quadrants all display
absent from beneath SW Texas and SE New
the correct moveout for Pds arrivals (Figure A1).
13 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
the data is divided into two random and disjoint
halves (Figure A2) and then restacked. This data
division analysis shows that a random half of the
total data set is sufficient to constrain the wave-
shape of the Pds arrivals. The plots for other data
bins (not shown) are similar.
A2. Source Deconvolution and Filtering
Effects
[42] To account for the filtering effects associated
with the deconvolution and the postdeconvolution
Figure A1. Pds moveout/phasing analysis for quadrant
stack bins NW‐B and SE‐B. Each radial receiver func-
tion is mapped from time to depth with the 1‐D veloc-
ity model described in the Data and Culling section.
The unity slope gray lines show where the phasing
depth equals the mapped depth, and the amplitude of
P to S conversions should be maximal. The solid black
lines are the Pds moveout corrected traces (same as
Figure 3). The red pluses correspond to the maximal
phasing amplitude. The results demonstrate that the
410‐LVL, 410, and 660 km discontinuity arrivals dis-
play the correct Pds moveout for these two selected
bins.
Figure A2. Random‐half receiver function quadrant
stacks for bins NW‐C, SE‐C, and SW‐C. The blue trace
is the mean receiver function stack, and the two disjoint
Sources from the SW back azimuth are not well random data halves are plotted as red and green lines.
distributed with respect to distance (Figure 2), and These representative examples demonstrate that the
thus this quadrant does not permit moveout anal- 410‐LVL, 410, and 660 km discontinuity signals are
ysis. To test the robustness of our quadrant stacks, robustly isolated by disjoint halves of the data set.
14 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
filter, the mean estimated impulse response is
calculated by stacking all deconvolved and filtered
vertical component receiver functions (Figure A3).
The first side lobes are 20% of the main peak
amplitude. This mean impulse response is con-
volved with our synthetic seismograms to facili-
tate direct comparison with the observed data.
A3. Marginal Posterior Probability
Density Function Integration
[43] To test whether the observed receiver function
data are normally distributed, normal cumulative
probability plots are calculated for the peak times
of the 410 km discontinuity and 410‐LVL arrivals
(Figure A4). These plots show that within nearly
two standard deviations of the mean (10% and 90%
probability on the ordinate) the data are normally
distributed. In contrast, beyond 10% and 90%
normal probability, the data contains “fat” proba-
bility tails with respect to a theoretical Gaussian
tail. These fat tails will thus make our standard
deviation estimates greater than that predicted by a
true normal distribution. With this caveat noted, we
conclude that our data are sufficiently normally
distributed to warrant the use of the L2 norm to
quantify misfit.
Figure A4. Data normality test. A Gaussian probabili-
ty distribution was fit to data samples at the peaks of the
discontinuity arrivals. (a) The 410 km discontinuity.
(b) The 410‐LVL. If the data were normally distributed,
the data would plot along the red dashed line. The slope
of the red line is proportional to the standard deviation
of the data, and the data value at which the line crosses
0.5 probability is the Gaussian mean of the data. The data
follow a normal distribution between approximately
10%–90% probability, after which the distribution has
fat tails with respect to a Gaussian distribution.
Acknowledgments
[44] The authors would like to acknowledge two anonymous
reviewers for thoughtful reviews and constructive comments,
which improved the manuscript. We also acknowledge
the facilities of the IRIS Data Management System, and
specifically the IRIS Data Management Center, for access to
the waveform and metadata required in this study.
References
Figure A3. Mean stacked waveform of the vertical re-
ceiver functions. This wavelet is convolved with the Bedle, H., and S. van der Lee (2009), S velocity variations
synthetic seismograms to account for the filtering and beneath North America, J. Geophys. Res., 114, B07308,
deconvolution effects in the processed data. doi:10.1029/2008JB005949.
15 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
Bercovici, D., and S. Karato (2003), Whole‐mantle convection Gu, Y. J., and A. M. Dziewonski (2002), Global variability of
and the transition‐zone water filter, Nature, 425, 39–44, transition zone thickness, J. Geophys. Res., 107(B7), 2135,
doi:10.1038/nature01918. doi:10.1029/2001JB000489.
Bolfan‐Casanova, N., S. J. Mackwell, H. Keppler, C. A. Helffrich, G. (2006), Extended‐time multitaper frequency do-
McCammon, and D. C. Rubie (2002), Pressure dependence main cross‐correlation receiver‐function estimation, Bull.
of H solubility in magnesiowüstite up to 25 GPa: Implications Seismol. Soc. Am., 96(1), 344–347, doi:10.1785/0120050098.
for the storage of water in the Earth’s lower mantle, Geophys. Helffrich, G. R., and B. J. Wood (2001), The Earth’s mantle,
Res. Lett., 29(10), 1449, doi:10.1029/2001GL014457. Nature, 412(6846), 501–507, doi:10.1038/35087500.
Bolfan‐Casanova, N., H. Keppler, and D. C. Rubie (2003), Hirschmann, M. M. (2006), Water, melting, and the deep Earth
Water partitioning at 660 km depth and evidence for very low H2O cycle, Annu. Rev. Earth Planet. Sci., 34, 629–653,
water solubility in magnesium silicate perovskite, Geophys. doi:10.1146/annurev.earth.34.031405.125211.
Res. Lett., 30(17), 1905, doi:10.1029/2003GL017182. Hirschmann, M. M., C. Aubaud, and A. C. Withers (2005),
Bolfan‐Casanova, N., C. A. McCammon, and S. J. Mackwell Storage capacity of H2O in nominally anhydrous minerals
(2006), Water in transition zone and lower mantle minerals, in the upper mantle, Earth Planet. Sci. Lett., 236(1–2),
in Earth’s Deep Water Cycle, Geophys. Monogr. Ser., 167–181, doi:10.1016/j.epsl.2005.04.022.
vol. 168, edited by S. D. Jacobsen and S. van der Lee, Hirschmann, M. M., A. C. Withers, and C. Aubaud (2006),
pp. 57–68, AGU, Washington, D. C. Petrologic structure of a hydrous 410 km discontinuity,
Burdick, S., C. Li, V. Martynov, T. Cox, J. Eakins, T. Mulder, in Earth’s Deep Water Cycle, Geophys. Monogr. Ser.,
L. Astiz, F. L. Vernon, G. L. Pavlis, and R. D. van der Hilst vol. 168, edited by S. D. Jacobsen and S. van der Lee,
(2008), Upper mantle heterogeneity beneath North America pp. 277–287, AGU, Washington, D. C.
from travel time tomography with global and USArray trans- Inoue, T., J. B. Parise, D. J. Weidner, and P. A. Northrup
portable array data, Seismol. Res. Lett., 79(3), 384–392, (1998), Elastic properties of hydrous ringwoodite (gamma
doi:10.1785/gssrl.79.3.384. phase) in Mg2SiO4, Earth Planet. Sci. Lett., 160(1–2),
Burdick, S., R. D. van der Hilst, F. L. Vernon, V. Martynov, 107–113, doi:10.1016/S0012-821X(98)00077-6.
T. Cox, J. Eakins, T. Mulder, L. Astiz, and G. L. Pavlis Inoue, T., K. Kojima, and T. Irifune (2007), Water content of
(2009), Model update December 2008: Upper mantle hetero- magma generated just above the 410‐km discontinuity, Eos
geneity beneath North America from P wave travel time to- Trans. AGU, 88(52), Fall Meet. Suppl., Abstract V24A‐06.
mography with global and USArray transportable array Irifune, T., et al. (1998), The postspinel phase boundary in
data, Seismol. Res. Lett., 80(4), doi:10.1785/gssrl.80.4.638. Mg2SiO4 determined by in situ X‐ray diffraction, Science,
Courtier, A. M., and J. Revenaugh (2006), A water‐rich tran- 279(5357), 1698–1700, doi:10.1126/science.279.5357.1698.
sition zone beneath the eastern United States and Gulf of Jasbinsek, J., andK. G. Dueker (2007), Ubiquitous low‐velocity
Mexico from multiple ScS reverberations, in Earth’s Deep layer atop the 410‐km discontinuity in the northern Rocky
Water Cycle, Geophys. Monogr. Ser., vol. 168, edited by Mountains, Geochem. Geophys. Geosyst., 8, Q10004,
S. D. Jacobsen and S. van der Lee, pp. 181–193, AGU, doi:10.1029/2007GC001661.
Washington, D. C. Jones, C. H., J. R. Unruh, and L. J. Sonder (1996), The role of
Dueker, K. G., and A. F. Sheehan (1997), Mantle discontinuity gravitational potential energy in active deformation in the
structure from midpoint stacks of converted P and S waves southwestern United States, Nature, 381(6577), 37–41,
across the Yellowstone hotspot track, J. Geophys. Res., doi:10.1038/381037a0.
102(B4), 8313–8327, doi:10.1029/96JB03857. Karato, S., D. Bercovici, G. Leahy, G. Richard, and Z. Jing
Efron, B., and R. J. Tibshirani (1994), An Introduction to the (2006), The transition‐zone water filter model for global
Bootstrap, Chapman and Hall, New York. material circulation: Where do we stand?, in Earth’s Deep
Fee, D., and K. Dueker (2004), Mantle transition zone topogra- Water Cycle, Geophys. Monogr. Ser., vol. 168, edited by
phy and structure beneath the Yellowstone hotspot, Geophys. S. D. Jacobsen and S. van der Lee, pp. 289–313,
Res. Lett., 31, L18603, doi:10.1029/2004GL020636. AGU, Washington, D. C.
Flanagan, M. P., and P. M. Shearer (1998a), Topography on Katsura, T., et al. (2004), Olivine‐wadsleyite transition in the
the 410‐km seismic velocity discontinuity near subduction system (Mg, Fe)2SiO4, J. Geophys. Res., 109, B02209,
zones from stacking sS, sP, and pP precursors, J. Geophys. doi:10.1029/2003JB002438.
Res., 103(B9), 21,165–21,182. Kohlstedt, D. L., H. Keppler, and D. C. Rubie (1996), Solubil-
Flanagan, M. P., and P. M. Shearer (1998b), Global mapping ity of water in the ALPHA, BETA and GAMMA phases of
of topography on transition zone velocity discontinuities by (Mg, Fe)2SiO4, Contrib. Mineral. Petrol., 123, 345–357,
stacking SS precursors, J. Geophys. Res., 103(B3), 2673–2692, doi:10.1007/s004100050161.
doi:10.1029/97JB03212. Kreutzmann, A., G. Marquart, I. T. Bjarnason, H. Schmeling,
Frost, D. J., and D. Dolejs (2007), Experimental determination A. Junge, and T. Ruedas (2004), Temperature and melting of
of the effect of H2O on the 410‐km seismic discontinuity, a ridge‐centred plume with application to Iceland. Part II:
Earth Planet. Sci. Lett., 256(1–2), 182–195, doi:10.1016/j. Predictions for electromagnetic and seismic observables,
epsl.2007.01.023. Geophys. J. Int., 159(3), 1097–1111, doi:10.1111/j.1365-
Futterman, W. I. (1962), Dispersive body waves, J. Geophys. 246X.2004.02397.x.
Res., 73, 3917–3935. Lawrence, J. F., and P. M. Shearer (2006), A global study of
Gao, W., E. Matzel, and S. P. Grand (2006), Upper mantle transition zone thickness using receiver functions, J. Geophys.
seismic structure beneath eastern Mexico determined from Res., 111, B06307, doi:10.1029/2005JB003973.
P and S waveform inversion and its implications, J. Geophys. Leahy, G. M., and D. Bercovici (2007), On the dynamics of a
Res., 111, B08307, doi:10.1029/2006JB004304. hydrous melt layer above the transition zone, J. Geophys.
Grand, S. P., and D. V. Helmberger (1984), Upper mantle Res., 112, B07401, doi:10.1029/2006JB004631.
shear structure of North America, Geophys. J. R. Astron. Lee, C.‐T. A., and W.‐P. Chen (2007), Possible density segre-
Soc., 76, 399–438. gation of subducted oceanic lithosphere along a weak serpen-
16 of 17
Geochemistry
Geophysics G3 JASBINSEK ET AL.: HYDRATED MANTLE BENEATH NEW MEXICO 10.1029/2009GC002836Geosystems
tinite layer and implications for compositional stratification of Smyth, J. R., and D. J. Frost (2002), The effect of water on the
the Earth’s mantle, Earth Planet. Sci. Lett., 255(3–4), 357– 410‐km discontinuity: An experimental study, Geophys. Res.
366, doi:10.1016/j.epsl.2006.12.022. Lett., 29(10), 1485, doi:10.1029/2001GL014418.
Litasov, K. D., E. Ohtani, and A. Sano (2006), Influence of Song, T. A., and D. V. Helmberger (2007), P and S waveform
water onmajor phase transitions in the Earth’s mantle, Earth’s modeling of continental sub‐lithospheric detachment at the
Deep Water Cycle, Geophys. Monogr. Ser., vol. 168, edited eastern edge of the Rio Grande Rift, J. Geophys. Res.,
by S. D. Jacobsen and S. van der Lee, pp. 95–111, AGU, 112, B07319, doi:10.1029/2007JB004942.
Washington, D. C. Song, T. A., D. V. Helmberger, and S. P. Grand (2004), Low
Liu, K. H. (2003), Effects of inelasticity on the apparent depth velocity zone atop the 410 seismic discontinuity in the north-
and detectibility of seismic discontinuities in the mantle,Geo- western United States, Nature, 427, 530–533, doi:10.1038/
phys. Res. Lett., 30(9), 1455, doi:10.1029/2002GL015264. nature02231.
Matsukage, K. N., Z. Jing, and S. Karato (2005), Density of hy- Thomas, C., and M. I. Billen (2009), Mantle transition zone
drous silicate melt at the conditions of the Earth’s deep upper structure along a profile in the SW Pacific: Thermal and
mantle, Nature, 438, 488–491, doi:10.1038/nature04241. compositional variations, Geophys. J. Int., 176, 113–125,
Murakami, M., K. Hirose, H. Yurimoto, S. Nakashima, and doi:10.1111/j.1365-246X.2008.03934.x.
N. Takafuji (2002), Water in Earth’s lower mantle, Science, Toffelmier, D. A., and J. A. Tyburczy (2007), Electromag-
295, 1885–1887, doi:10.1126/science.1065998. netic detection of a 410‐km‐deep melt layer in the south-
Ohtani, E., H. Mizobata, and H. Yurimoto (2000), Stability western United States, Nature, 447(7147), 991–994,
of dense hydrous magnesium silicate phases in the sys- doi:10.1038/nature05922.
tems Mg2SiO4–H2O and MgSiO3–H2O at pressures up to Turcotte, D. L., and G. Schubert (2002), Geodynamics, 2nd
27 GPa, Phys. Chem. Miner., 27, 533–544, doi:10.1007/ ed., Cambridge Univ. Press, Cambridge, U. K.
s002690000097. van der Lee, S., H. Paulssen, and G. Nolet (1994), Variability
Park, J. (1987), Multitaper spectral analysis of high‐frequency of P660s phases as a consequence of topography of the
seismograms, J. Geophys. Res., 92(B12), 12,675–12,684, 660 km discontinuity, Phys. Earth Planet. Inter., 86(1–3),
doi:10.1029/JB092iB12p12675. 147–164, doi:10.1016/0031-9201(94)05066-X.
Park, J. (1996), Surface waves in layered anisotropic struc- Vinnik, L., and V. Farra (2007), Low S velocity atop the
tures, Geophys. J. Int., 126, 173–184, doi:10.1111/j.1365- 410‐km discontinuity and mantle plumes, Earth Planet. Sci.
246X.1996.tb05276.x. Lett., 262(3–4), 398–412, doi:10.1016/j.epsl.2007.07.051.
Park, J., and V. Levin (2000), Receiver functions from multi‐ Vinnik, L., M. R. Kumar, R. Kind, and V. Farra (2003), Super‐
taper spectral correlation estimates, Bull. Seismol. Soc. Am., deep low‐velocity layer beneath the Arabian plate, Geophys.
90(6), 1507–1520, doi:10.1785/0119990122. Res. Lett., 30(7), 1415, doi:10.1029/2002GL016590.
Revenaugh, J., and S. A. Sipkin (1994), Seismic evidence for Vinnik, L. P., V. Farra, and R. Kind (2004), Deep structure of
silicate melt atop the 410‐km mantle discontinuity, Nature, the Afro‐Arabian hotspot by S receiver functions, Geophys.
369(6480), 474–476, doi:10.1038/369474a0. Res. Lett., 31, L11608, doi:10.1029/2004GL019574.
Roth, J. B., M. J. Fouch, D. E. James, and R. W. Carlson Wilson, D., and R. Aster (2005), Seismic imaging of the crust
(2008), Three‐dimensional seismic velocity structure of the and upper mantle using regularized joint receiver functions,
northwestern United States, Geophys. Res. Lett., 35, frequency–wave number filtering, and multimode Kirchhoff
L15304, doi:10.1029/2008GL034669. migration, J. Geophys. Res., 110, B05305, doi:10.1029/
Sakamaki, T., A. Suzuki, and E. Ohtani (2006), Stability of hy- 2004JB003430.
drous melt at the base of the Earth’s upper mantle, Nature, Wilson, D., S. Grand, W. Gao, W. S. Baldridge, S. Semken,
439(7073), 192–194, doi:10.1038/nature04352. P. Patel, R. Aster, M. West, and J. Ili (2005a), Lithospheric
Sambridge, M. (1999), Geophysical inversion with a Neigh- structure of the Rio Grande rift, Nature, 433(7028),
bourhood Algorithm—II, Geophys. J. Int., 138, 727–746, 851–855, doi:10.1038/nature03297.
doi:10.1046/j.1365-246x.1999.00900.x. Wilson, D., M. West, W. Gao, W. S. Baldridge, S. Semken,
Shen, Y., and J. Blum (2003), Seismic evidence for accumulated R. Aster, J. Ni, and S. Grand (2005b), Imaging the seismic
oceanic crust above the 660‐km discontinuity beneath south- structure of the crust and upper mantle beneath the Great
ern Africa, Geophys. Res. Lett., 30(18), 1925, doi:10.1029/ Plains, Rio Grande Rift, and Colorado Plateau using receiver
2003GL017991. functions, J. Geophys. Res., 110, B05306, doi:10.1029/
Shieh, S. R., H. Mao, R. J. Hemley, and L. C. Ming (1998), 2004JB003492.
Decomposition of phase D in the lower mantle and the fate Wood, B. J. (1995), The effect of H2O on the 410‐kilometer
of dense hydrous silicates in subducting slabs, Earth Planet. seismic discontinuity, Science, 268, 74–76, doi:10.1126/
Sci. Lett., 159, 13–23, doi:10.1016/S0012-821X(98)00062-4. science.268.5207.74.
Sine, C. R., D. Wilson, W. Gao, S. P. Grand, R. Aster, J. Ni, Youngs, B. A. R., and D. Bercovici (2009), Stability of a
and W. S. Baldridge (2008), Mantle structure beneath the compressible hydrous melt layer above the transition zone,
western edge of the Colorado Plateau, Geophys. Res. Lett., Earth Planet. Sci. Lett., 278, 78–86, doi:10.1016/j.
35, L10303, doi:10.1029/2008GL033391. epsl.2008.11.024.
17 of 17