Article
https://doi.org/10.1038/s41467-024-47887-0
Seismic evidence for melt-rich lithosphere-
asthenosphere boundary beneath young
slab at Cascadia
Xin Wang
1,2,3
,LingChen
2,4
,KelinWang
5
,Qi-FuChen
1,2
,
Zhongwen Zhan
3
&JianfengYang
2,4
The Lithosphere-Asthenosphere Boundary (LAB) beneath oceanic plates is
generally imaged as a sharp seismic velocity reduction, suggesting the pre-
sence of partial melts. However, the fate of a melt-rich LAB is unclear after
these plates descend into the mantle at s
ubduction zones. Recent geophysical
studies suggest its persistence with do
wn-going old and cold slabs, but whe-
ther or not it is commonly present remains unclear, especially for young and
warm slabs such as in the Cascadia subduction zone. Here we provide evidence
for its presence at Cascadia in the form of a large (9.8 ± 1.5%) decrease in shear-
wave velocity over a very small (<3 km) depth interval. Similarly large and
sharp seismic velocity reduction at t
he bottom of both old and young slabs, as
well as along the base of oceanic plates before subduction, possibly represents
widespread presence of melts. The m
elt-rich sub-slab LAB may strongly
in
fl
uence subduction dynamics and vis
coelastic earthquake cycles.
The Lithosphere-Asthenosphere Boundary (LAB) beneath oceanic
plates marks a sharp decrease in seismic velocity. The sharp decrease,
loosely but commonly referred to as a negative
“
seismic discontinuity
”
,
is widely thought to indicate the presence of partial melts
1
–
5
,although
there are competing interpretations
6
,
7
. A melt-rich LAB may mechani-
cally decouple the lithospheric plate from the underlying astheno-
sphere and thus facilitate plate motion
8
,
9
. If it continues into
subduction zones, similar decoupling would occur beneath the sub-
ducting plate (called the slab) and fundamentally in
fl
uence subduction
dynamics
10
and viscoelastic earthquake cycles
11
. Here we refer to the
LAB of an oceanic plate before subduction as the plate-LAB and
that after subduction as the slab-LAB (Fig.
1
). However, it is by no
means clear whether the slab-LAB is as common as the plate-LAB. The
slab-LABhasbeenwellimagedintheJapanandNewZealandsub-
duction zones
1
,
8
, and in both cases, the age of the subducting plate is
very old
12
(~120
–
130 Ma; Fig.
1
a). In contrast, its presence in regions
where young plates are subducting remains unclear
3
.Amagneto-
telluric study off the Middle America trench could resolve a melt-rich
plate-LAB beneath the young (~23 Ma) Cocos plate but not a slab-LAB
in the subduction zone
13
. The plate-LAB is seismically detected beneath
almost the entire young (<10 Ma) Juan de Fuca plate, but so far, there is
no reported evidence for the slab-LAB in the Cascadia subduction
zone
14
. Whether or not the presence of a melt-rich slab-LAB depends
on slab age is a question with important geodynamic implications.
Cascadia is an end-member warm-slab subduction zone where the
incoming Juan de Fuca plate is young (<10 Ma) and warm (Fig.
1
). Using
teleseismic converted waves generated at seismic discontinuities
(referred to as receiver functions, RFs), previous studies reported
various discontinuities related to the subducting Juan de Fuca slab,
including those associated with the
fl
uid-rich subducting oceanic
crust, slab-Moho, and anisotropy within the slab crust and/or
mantle
15
–
22
. However, no seismic discontinuity associated with slab-
LAB has been reported thus far. Seaward of the subduction zone,
Rychertetal.detected
14
the plate-LAB throughout the incoming Juan
de Fuca plate and considered it strong evidence for the presence of
partial melts along the base of the plate. However, whether the inferred
Received: 13 December 2023
Accepted: 15 April 2024
Check for updates
1
Key Laboratory of Earth and Planetary Physics, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China.
2
College of Earth and
Planetary Sciences, University of Chinese Academy of Sciences, Beijing, China.
3
Seismological Laboratory, California Institute of Technology, Pasadena, CA,
USA.
4
State Key Laboratory of Lithospheric Evolution, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing, China.
5
Paci
fi
c Geoscience
Centre, Geological Survey of Canada, Sidney, BC, Canada.
e-mail:
lchen@mail.iggcas.ac.cn
Nature Communications
| (2024) 15:3504
1
1234567890():,;
1234567890():,;
melt-rich LAB persists after the plate is subducted remains an unre-
solved question.
To address this question, we examine P-to-S (Ps conversion) RFs
by taking advantage of dense seismic arrays at Cascadia (Fig.
1
). One
array consists of broadband seismometers with a station spacing of
~5kmandoperatedin1993
–
1994; the other consists of short-period
nodal seismometers with a station spacing of ~500 m and operated for
one month in 2017 (Supplementary Fig. 1). Data from these arrays have
been used in previous studies for investigating the shallow sedimen-
tary basin structures, mid-crustal Conrad discontinuity, continental
Moho, slab geometry,
fl
uid-rich slab crust and its dehydration pro-
cesses, and fossil anisotropy within the slab mantle
15
,
17
,
20
–
26
.Belowthe
slab-Moho, some patchy negative RF signals have been detected
previously
21
,
22
,
24
,
25
, which might be indicative of the presence of a slab-
LAB. However, these studies did not consider them to be structural
signals because of their intermittent appearance and concerns about
artifacts arising from the side-lobe effect of RF deconvolution
and multiple reverberations from shallow structures. One study
21
tentatively attributed these negative signals to anisotropy within a
very thick slab, yet their
fi
ndings might be in
fl
uenced by cross-mode
132 ̊W
130 ̊W
128 ̊W
126 ̊W
124 ̊W
122 ̊W
44 ̊N
46 ̊N
48 ̊N
200 km
20 km
40 km
200 km
Portland
Seattle
Juan de
Fuca plate
45 mm/yr
Ridge
20 km
20 km
40 km
60 km
XZ Broad-band stations
ZO Short-period stations
Slab-LAB
Plate-LAB
S
u
b
d
u
c
t
i
n
g
s
l
a
b
Slab Moho
Slab
top
?
Continental
Moho
?
Age of Oceanic lithosphere (Ma)
0
20
40
60
80
100
120
140
160
180
a
b
c
Fig. 1 | Study area and illustration of the seismic discontinuities discussed in
this study. a
Global map showing the age distribution of the oceanic lithosphere
with data from ref.
12
. Our study area (star), the Cascadia subduction zone, has a
young and warm slab, in contrast to the old and cold slab in the other two sub-
duction zones (squares; Japan and New Zealand) where evidence for sub-slab melt-
rich Lithosphere-Asthenosphere Boundary (LAB) has been reported
1
,
8
.
b
Regional
tectonic setting of our study area and locations of seismic stations used in this
study. Dashed lines show depth to the surface of the slab from Slab2.0 model
59
.The
maps were generated using Generic Mapping Tools
60
, with topography and
bathymetry data from the Global Multi-Resolution Topography Synthesis
61
.
c
Cartoon showing how the LAB beneath the Juan de Fuca plate (plate-LAB, based
on ref.
14
) may continue after subduction to give rise to the slab-LAB. Question
marks indicate that the origination of melts for the LAB remains under debate, and
the depth extent of the slab-LAB requires further investigation.
Article
https://doi.org/10.1038/s41467-024-47887-0
Nature Communications
| (2024) 15:3504
2
artifacts resulting from the usage of a multiple-mode conversion
imaging technique
20
. In our present study, by employing a recently
developed Bayesian array-based coherent receiver function (CRF)
imaging technique
27
,
28
, and conducting multiple frequency analysis
and extensive synthetic tests, we are able to detect the slab-LAB from
these data in the form of strong and sharp seismic velocity reduction.
We consider the imaged slab-LAB to be compelling evidence for the
continuation of the melt-rich plate-LAB after subduction.
Results
CRF imaging and comparison with conventional RF analyses
Inthedistancerange(30°
–
90°) optimal for CRF analyses, the arrays
recorded seismic waves generated by distant earthquakes mainly in
the southeast, northwest, and southwest directions (Supplementary
Fig. 1). Because the Juan de Fuca slab dips to the east, the Ps waves from
the southeast are more effective than those from the other directions
in constraining slab-parallel seismic discontinuities (Supplementary
Figs. 2, 3). We, therefore, mainly use the Ps waves from this direction
for our analyses. The results based on waves from the other directions
provide corroborative information. We process the data in multiple
frequency bands (0.05
–
0.5 Hz, 0.15
–
1.0 Hz, 0.25
–
1.5 Hz, 0.35
–
2.0 Hz)
and employ the CRF technique to conduct a joint analysis of the
broadband and short-period data (see Methods; Supplementary
Figs. 4
–
6). The denser spatial coverage of the short-period array allows
us to better constrain shallow structures, which aids in constraining
deep structures (see Methods).
Compared to previous RF studies
15
,
22
,
26
, which predominantly
focused on relatively low-frequencies (up to 0.3 Hz), we use multiple
frequency analyses mentioned above to check the persistence of dis-
continuities. The frequency dependence of the amplitudes of Ps pha-
ses helps to appraise the robustness and sharpness of the
discontinuities
29
,
30
. Although increasing the frequency content of the
data increases the vertical resolution of subsurface structures, high-
frequency analysis is prone to noise and local scattering. To overcome
this challenge, we employ the array-based CRF technique, which can
leverage the coherency of the wave
fi
eld recorded by a dense array to
suppress incoherent noise and local scattering (Fig.
2
). The CRF
method is also less susceptible to interference to the target dis-
continuity image from nearby strong discontinuities, as shown by the
comparison of subsurface images obtained using the CRF and con-
ventional RF methods (Fig.
2
and Supplementary Fig. 7). Another
advantage of the CRF is that the reliability of imaged structures is
measured using probability distribution, allowing objective assess-
ment of structure identi
fi
cation and interpretation
28
.
The CRF images are originally obtained in the time domain. To
convert time-to-depth, we employ the seismic velocity model from the
tomography study of ref.
31
. Although the spatial resolution of this
model is much higher than other available regional tomography
models
32
–
34
(Supplementary Fig. 8), its velocity
fi
eld is still inevitably
damped by the smoothing scheme in tomographic inversion and thus
distorts the depths of discontinuities in the CRF imaging (Supple-
mentary Figs. 9, 10). As will be detailed in the following section, cor-
rection for this distortion is an integral part of the interpretation of our
CRF results. The CRF image based on seismic waves from the southeast
is shown in Fig.
3
. Seismic waves from the northwest are theoretically
less optimal for resolving east-dipping discontinuities (Supplementary
Fig. 2), but they still result in comparable CRF images for an area
immediately to the north (Supplementary Fig. 11), offering further
support to the main imaged features shown in Fig.
3
. Because of the
potential depth distortion discus
sed above, we refer to the depths
shown in Fig.
3
as apparent depths.
The slab-top and slab-Moho discontinuities
Shallower than 60 km, our CRF image reproduces several primary
structure boundaries found in previous studies
16
,
20
,
21
,
25
,
35
and thus
demonstrates the ef
fi
cacy of the CRF technique. At 0
–
10 km apparent
depths, prominent positive discontinuities (Fig.
3
a) roughly delineate
known sedimentary basins
26
,
35
. The negative intra-crustal discontinuity
observed at about 10
–
20 km depth has also been observed by previous
studies
20
,
25
and may be related to the accumulation of
fl
uids within the
continental crust
36
,
37
.Inthe20
–
50 km range, two subparallel east-
dipping seismic discontinuities with opposite polarity delineate the
slab geometry and the low-velocity subducting crust. By utilizing
higher-frequency data in comparison to previous studies
15
,
24
,ourpro-
posed slab geometry contains more details; however, it is important to
note that the tomographic model used in time-to-depth conversion
may introduce slight distortions to the slab geometry, both in our
study and in previous ones. The apparent thickness of the low-velocity
zone >10 km is larger than what is expected for a normal oceanic crust
due to the aforementioned depth distortion caused by velocity
smoothing in the tomography model that is used for time-depth
conversion. If we follow previous studies in Cascadia to assume a
fl
uid-
saturated crust with a very low Vs (about 20
–
50% Vs reduction) and a
high Vp/Vs ratio (about 2
–
3)
16
,
17
, the thickness is corrected to be
~7
–
8 km (Supplementary Fig. 10). The CRF amplitudes and our syn-
thetic tests indicate extremely large velocity contrasts across the top
(~15
–
20%) and bottom (~35
–
55%) of the subducting crust (see details in
Methods, Fig.
4
), consistent with previous estimates
16
,
17
. The decrease
in the amplitudes of the Ps conversions at the slab-top and slab-LAB
when the slab depth exceeds ~40 km (Fig.
3
a) is likely due to the
eclogitization process of the subducting oceanic crust
15
,
23
. Eclogitiza-
tion is accompanied by dehydration and thus releases aqueous
fl
uid
into the overlying forearc mantle wedge, leading to the well-known
absence of a detectable continental Moho near the mantle wedge
corner
15
.
The slab-LAB discontinuity
The most important
fi
nding enabled by the newly employed CRF
technique is the continuous negative discontinuity at about 60
–
80 km
apparent depths, which we interpret as the slab-LAB (Fig.
3
). A direct
estimate of the actual depth of the negative discontinuity using
observed Ps arrival times show that it is located about 25 km below the
slab-Moho or about 32 km below the slab-top (Supplementary Fig. 10).
Patchy negative signals at relevant depths can be seen in some other
seismic imaging studies at Cascadia
20
,
21
,
24
,
25
, but it was not clear whether
they were structural signals or imaging artifacts. Here, we are able to
resolve con
fi
dently this discontinuity owing to a combination of mul-
tiple frequency analyses, the narrow back-azimuth range of earth-
quakes used in dipping discontinuity imaging, and the advantages of
the CRF method. The Bayesian posterior probability distribution of the
CRF phases shows a high con
fi
dence level for this discontinuity
(Fig.
3
b), and its presence is persistent across multiple frequency
bands (Supplementary Figs. 5, 6).
Our synthetic tests verify that the strong negative signal marking
this discontinuity cannot be an imagin
g artifact due to other structures
(Supplementary Figs. 12
–
16). Basin reverberations occur at shallower
depths, and multiples from the slab-top occur at much deeper depths
(Supplementary Fig. 12). Multiples from a positive intra-crustal dis-
continuity (i.e., the Conrad discontinuity) at ~15 km depth in the
overriding continental plate would exhibit a pair of positive and
negative signals at a depth similar to the observed negative dis-
continuity (Supplementary Fig. 13). However, such a Conrad dis-
continuity is not observed in our CRF image (Fig.
3
)andthe
conventional RF images (Supplementary Fig. 5). Although a negative
intra-crustal discontinuity is imaged at a depth of 10
–
15 km in our
study, it generates multiples with ge
ometry inconsistent with the slab-
LAB discontinuity (Supplementary Figs. 13, 14). More importantly, the
RF images constructed with seismic waves from the northwest show a
weaker negative signal than that obtained using the waves from the
southeast (Supplementary Figs. 3, 15), and the arrival time of the
Article
https://doi.org/10.1038/s41467-024-47887-0
Nature Communications
| (2024) 15:3504
3
negative phase decreases with the increase of epicentral distance
(Supplementary Fig. 16), suggesting that the negative signal is related
to the direct Ps phase rather than multiples. We further analyze the RFs
as a function of back-azimuth using both synthetic data and real
observations, and our results show that this negative signal is better
explained by a dipping discontinuity rather than an anisotropy layer
(Supplementary Figs. 17
–
19).
Sharpness and velocity reduc
tion across the slab-LAB
An important attribute of the inferred LAB discontinuity is its sharp-
ness, characterized by the magnitude of the velocity change (dVp or
dVs) and the depth range (dZ) over which the change occurs (Fig.
4
).
The amplitude of the Ps phase in RFs is particularly sensitive to dVs, dZ,
and the range of dominant frequencies used in the analysis. If dZ is
broader than half of the dominant wavelength of the incident P-wave,
the amplitude decreases dramatically with increasing frequency
8
,
29
(Supplementary Fig. 20). The Ps conversions of our inferred slab-LAB
consistently exhibit an amplitude of 12 ± 3% of the direct P-arrival
across the entire frequency range (Fig.
4
). Assuming a typical upper-
mantle Vp of ~8 km/s and considering the waveform frequencies used
in this study up to 2.0 Hz, the dZ related to this discontinuity should
not much exceed ~2 km
29
. Further synthetic tests indicate that the slab-
LAB corresponds to a sharp discontinuity with dVs of 9
:
8±1
:
5% over a
depth range of 1
:
5±1
:
5km(Fig.
4
and Supplementary Fig. 21). Such a
sharp velocity reduction within 3 km cannot be solely explained by
variations in water abundance, temperature, or grain size
1
,
3
,
38
.Arecent
study
39
suggests that hydration within the asthenosphere can sub-
stantially reduce seismic velocities, offering a potential explanation for
the strong and sharp velocity reduction across the LAB. However, the
effect of hydration on velocity reduction becomes less pronounced for
seismic waveform frequencies considered in this study. Thus, the
presence of partial melts stands out as the most likely mechanism for
120
100
80
60
40
20
0
Depth (km)
0
5
10
15
20
25
Time (sec)
0
5
10
15
20
25
Time (sec)
0 102030405060708090100110120130140
Distance (km)
0 102030405060708090100110120130140
Distance (km)
-20
0
20
40
60
80
100
120
140
160
180
Sla
b top
Ps ph
ase
Slab M
oho
P
s
pha
se
S
l
ab_
LAB Ps pha
se
Multiple
reverberations
Conventional RF method
0.05 ~ 0.3 Hz
Conventional RF method
0.15 ~ 1.5 Hz
Array-based CRF method
0.15 ~ 1.5 Hz
Slab-LAB
Slab Moho
Slab top
a
c
e
0 102030405060708090100110120130140
0
5
10
15
20
25
0
5
10
15
20
25
0
5
10
15
20
25
Time (sec)
b
Array-based CRF method
0.05 ~ 0.3 Hz
d
Time (sec)
Time (sec)
f
RF amplitude
-0.2
-0.1
0.0
0.1
0.2
Multiple
reverberations
Conventional RF method
0.15 ~ 1.5 Hz
Noise free
Slab top Ps phase
Slab Moho Ps phase
Slab_LAB Ps phase
2.5
3.0
3.5
4.0
4.5
5.0
Vs(km/s)
Fig. 2 | Synthetic tests to compare the array-based coherent receiver function
(CRF) method with the conventional receiver function (RF) method. a
The
synthetic model based on the slab geometry in central Cascadia and the distribu-
tion of the broadband seismic stations (triangles) shown in Fig.
1
b. Synthetic seis-
mograms are generated for the speci
fi
c source-receiver pairs (with the earthquakes
located in the southeast direction in Supplementary Fig. 1) in central Cascadia to
simulate real situations.
b
Conventional RF stacking images for noise-free synthetic
data. In this scenario, only high-frequency (0.05
–
1.5 Hz) results are presented to
highlight RF signals related to slab-related discontinuities.
c
,
d
Conventional RF
stacking images for synthetic data with realistic noise. With low frequency
(0.05
–
0.3 Hz), the slab-top and slab-Moho can be coherently identi
fi
ed only with
their multiples, and it is dif
fi
cult to recognize the slab-LAB. With high frequency
(0.05
–
1.5 Hz), slab-related signals manifest as intermittent features.
e
,
f
Array-based
CRF stacking images for synthetic data with realistic noise. The slab-top, slab-Moho,
and slab-LAB can be resolved in both low- and high-frequency imaging. See Sup-
plementary Fig. 7 for noise-free synthetic tests.
Article
https://doi.org/10.1038/s41467-024-47887-0
Nature Communications
| (2024) 15:3504
4
0
10
20
30
40
50
60
70
80
90
100
Apparent Depth (km)
Apparent Depth (km)
0
10
20
30
40
50
60
70
80
90
100 110 120 130
Distance to coast (km)
140
RF amplitude
3.0
3.5
4.0
4.5
Vs (km/s)
a
c
Slab Moho
Slab-LAB
Slab top
Slab
top
Slab Moho
Slab-LAB
?
?
Apparent Depth (km)
b
Probability of CRF
0.6
0.7
0.8
0.9
1
0.2
0.3
0.4
0.5
0
0.1
Corrected slab-crust thickness ~7–8 km
Corrected slab-mantle thickness ~25 km
High
Low
Mag
3
2
1
-0.2
-0.1
0.0
0.1
0.2
0
10
20
30
40
50
60
70
80
0
10
20
30
40
50
60
70
80
90
100
Topography
Fig. 3 | Coherent receiver function (CRF) stacking image and interpretation.
a
CRF stacking image obtained by jointly inverting the broadband and short-period
data using teleseismic waves (0.15
–
1.0 Hz) from the southeast. Interpretation of the
three important discontinuities is as labeled. The apparent depths estimated for
these discontinuities suffer from distortion caused by smoothing in the seismic
velocity model employed for time-depth conversion. As discussed in the text and
shown in Supplementary Fig. 10, the actual separations between the slab-top and
slab-Moho and between the slab-Moho and slab-LAB are estimated to be ~7
–
8and
~25 km, respectively.
b
Density plot of the ensemble solutions represents the pos-
terior probability distribution of the CRF phases, which serves as a measure of the
reliability of the imaged structures.
c
Same three CRF discontinuities from (a) in the
backdrop of the tomographic image of ref.
31
used for our time-to-depth conver-
sion. Open circles show earthquakes from the Paci
fi
c Northwest Seismic Network
catalog within 10 km of the cross-section. In all the panels, the thin dashed and
dotted gray lines represent the subduction interface from Slab2.0 model
59
and
ref.
62
, respectively.
Article
https://doi.org/10.1038/s41467-024-47887-0
Nature Communications
| (2024) 15:3504
5
the strong and sharp velocity reduction at the slab-LAB. Similar to
previous estimates
1
,
3
,
14
,adVsof~9
:
8±1
:
5% across the LAB can be
attributed to the presence of ~1
–
4% melt fraction, depending on the
melt geometry
40
.
Discussion
Negative seismic discontinuities below the slab-Moho have been pre-
viously recognized at Cascadia, but they are too shallowly (about
5
–
15 km) located beneath the slab-Moho to be the slab-LAB. They are
thus inferred to indicate anisotropy of the subducting oceanic
uppermost mantle from fossil fabrics generated at spreading
ridges
18
,
21
. In contrast, our LAB discontinuity is located at a greater
depth and outlines the lower boundary of the high-velocity slab (Fig.
3
and Supplementary Fig. 11). As explained above, detailed comparisons
between real observations and synthetic data show that our imaged
structure is more consistent with a dipping feature than with aniso-
tropy. More importantly, its about 32 km depth below the top of the
slabiscompatiblewiththedepthoftheJuandeFucaplate-LABbefore
subduction (Supplementary Fig. 22). It is thus natural to consider the
slab-LAB imaged in this work a continuation of the plate-LAB.
It is widely assumed that partial melting at the uppermost asthe-
nosphere is responsible for the presence of a melt-rich plate-LAB
3
.Itis
important to ask whether the partial melts at the slab-LAB are gener-
ated in situ with a similar origin or instead re
fl
ect the downdip
transport of plate-LAB melts by the subducting slab. Hawley et al.
speculated
34
that the melt-rich plate-LAB material traveled laterally
with the Juan de Fuca plate towards the Cascadia subduction zone but
mainly piled up beneath the trench area because its buoyancy would
resist subduction
13
, giving rise to a tomographically detected large low-
velocity volume. However, in a later tomographic study, Bodmer et al.
did
41
not
fi
nd the low-velocity volume in central Cascadia, including
our study area, putting the notion of melt piling up and hence the
downdip transport scenario in question. Nonetheless, neither of these
tomographic studies has adequate vertical spatial resolution to resolve
the sharpness of the LAB discontinuity imaged in our study.
Considering the inference of melt-rich LAB with comparable melt
fractions beneath oceanic plates both before and after subduction
1
,
8
,
14
,
we speculate that the melts originate predominantly locally, without
long-distance lateral migration from ridges or plumes. For the Juan de
Fuca plate, Rychert et al. detected
14
a weaker and intermittent positive
discontinuity at ~90
–
130 km depths near the mantle solidus as well as
the plate-LAB at ~20
–
45 km depths. Partial melting may occur between
the two discontinuities, with melt ponding at the base of the less-
permeable oceanic lithosphere over geological timescales
42
,
43
to cause
the observed sharp velocity reduction. A typical slab thermal model in
Cascadia with a mildly hydrated mantle (water contents of 100
–
200
ppm) could also result in the occurrence of in situ partial melting to
explain the observed sharp slab-LAB in our study (Supplementary
-35
-30
-25
-20
-15
-10
-5
0
5
10
15
20
25
30
35
40
45
50
55
60
65
RF amplitude (%)
0
10
20
30
40
50
60
70
80
Frequency (%)
-40-30-20-10 0 1020304050607080
dVs (%)
0
10
20
30
40
50
60
Frequency (%)
0
5
10
15
20
25
dZ (km)
Bandpass filter
0.05-0.5 Hz
0.15-1.0 Hz
0.25-1.5 Hz
0.35-2.0 Hz
Slab top Slab Moho Slab-LAB
Obs.
Syn.
dZ (km)
0
4
8
12
16
20
24
dVs
dZ
Vs
Depth
Slab
LAB
Slab top
Slab Moho
c
b
a
Fig. 4 | Estimation of seismic velocity structures. a
Amplitude of the Ps phases as
a function of frequency. The observations are shown with colored circles as de
fi
ned
in the legend, with the error bar representing one standard deviation. The synthetic
counterparts are shown as lines color-coded by the depth range (dZ) over which
velocity decrease/increase occurs, as de
fi
ned in the legend. For display clarity, we
only show the models with a relative mis
fi
t(asde
fi
ned in Methods) of less than
100%.
b
,
c
Histograms showing dZ and the magnitude (dVs), respectively, of the
velocity decrease/increase associated with the three discontinuities.
Article
https://doi.org/10.1038/s41467-024-47887-0
Nature Communications
| (2024) 15:3504
6